Virsto Blog - Virsto Adding Citrix XenDesktop Support

Virsto Adding Citrix XenDesktop Support

by Eric Burgener, VP Product Management

Posted Wednesday, May 09, 2012 in News 0 comments

According to analyst firm Gartner, the VDI market is almost evenly split on ESX/ESXi between the leading virtual desktop front ends VMware View and Citrix XenDesktop, with each contender laying claim to about 40% of the market. With the recent announcement of our support for XenDesktop 5 on ESX/ESXi, we’re now offering Virsto’s performance, space-efficiency, cluster-awareness, and time-saving benefits to roughly 80% of the VDI market on vSphere.

In the same way that we support VMware View 4 and 5 on vSphere, we are integrating in with XenDesktop in a way that completely preserves the deployment, re-deployment, and refresh workflows that XenDesktop administrators are using in Citrix Desktop Studio and/or Desktop Director.

No re-training is required – we just transparently use Virsto during the cloning step instead of either PVS or MCS, almost instantly creating up to thousands of high performance, space-efficient, cluster-aware clones. This saves hours (and in some cases, days) on overall desktop provisioning times and, unlike host-based SSDs that you might be thinking about using to address storage performance issues, fully supports failover. Once the desktops are created, unique identities assigned, and placement into desktop pools is complete, the administrator manages them in EXACTLY the same way that non-Virsto XenDesktops would be managed.

With a simple software install and the purchase of NO additional storage hardware, you will find that Virsto will cut the storage cost/desktop in half, double the desktop density that you can support per host, and not require that you purchase additional memory or are forced to deploy with more expensive 4 socket motherboards. Our log architecture allows you to handle I/O spikes like those generated by boot, login, application, and logout storms without storage over-provisioning, which means that you’ll need at least 50% less storage to meet your virtual desktop deployment requirements.

With Virsto, many desktop performance requirements can be met without SSD. But if you choose to deploy VDI with SSD, Virsto’s unique logging architecture and tiered storage capacity delivers read and write acceleration as if the entire VDI environment was running on SSD, only using 95% less SSD capacity than an SSD-based cache. User profile data can be stored on cheaper SATA disk, while experiencing SSD performance.

If you’re looking into virtual desktops on vSphere, using either View or XenDesktop, I invite you to check us out. And if you have a VDI project that has stalled because of storage performance or cost issues, we can get that project back on track for you.

View 5.1’s Storage Accelerator a Great Complement To Virsto

by Eric Burgener, VP Product Management

Posted Thursday, May 03, 2012 in Technology 0 comments

Tags: Virsto VDI, VMware

Earlier this week, VMware announced the release of View 5.1 with a new Storage Accelerator feature. Since storage acceleration is part of the Virsto value proposition, a question that is likely to come up is what the effect of Storage Accelerator on the Virsto value proposition is. The net net is that Storage Accelerator does not diminish what Virsto offers in View environments and the two potentially complement each other very well.

From VMware’s own description, Storage Accelerator is a host-based, read caching approach that provides acceleration for desktop reads using RAM. VMware didn’t talk about their caching algorithms, but if it works like most caching algorithms, it will page in blocks based on frequency of access, and since it will do that in “chunks”, you’ll get other blocks near to the ones actually accessed that are likely to be accessed as well in the future. Caching algorithms by definition look “backwards” so you don’t get the performance the first time you access a block, only once that block has already been moved into cache.

In VDI environments, the golden master (the OS disk that includes the blocks that are shared across all the desktops) will get moved to cache pretty quickly, so you’ll get good performance there out of Storage Accelerator. This will help during read intensive operations like boot storms, and will also provide performance improvements whenever the golden master needs to be read. What it won’t do is address write performance.

If you’re familiar with the I/O patterns in VDI environments, then you know that they tend to be very write-intensive, much moreso even than virtual server environments. Write to read ratios will vary across environments, but its not unusual to have 80% writes and 20% reads during “steady state” VDI I/O. I attended the San Luis Obispo VMUG meeting on Monday, Apr 30, where VMware in a presentation suggested that the steady state planning number should be more like 90% writes, 10% reads. Because of this heavy skewing towards writes, write performance is much more of a problem in the storage slowdown that occurs in VDI environments, and the Storage Accelerator does nothing to address that.

Virsto’s log architecture, on the other hand, is perfect for boosting write performance using very little log capacity. In fact, the more random and the more write-intensive your environments are, the greater performance improvements you’ll see from using Virsto. You can use Virsto together with Storage Accelerator to comprehensively speed up both writes and reads, leveraging a mix of storage and memory to do so. Or, if you do not want to use memory for read speedups, you can create a very small tier 0 out of fast storage (like FC 15K RPM spinning disks or SSD) to get read speedups all the time for all the desktops that reside on it. The tier 0 is of course a great place to put golden masters.

There are a few considerations when looking at the memory vs disk decision. Main memory’s a bit expensive (you can buy 2TB of memory for a Dell PowerEdge R910 for about $75/GB, which is a little more than what you’d pay on a $/GB basis for enterprise class SSD), and if you’re going to create big memory-based caches, you’re probably going to have to buy a 4 socket server to accommodate the capacity you want. 4 socket servers are quite a bit more expensive than 2 socket ones (you’ll pay about $28K list for a Dell PowerEdge R910). The bigger your cache capacity, the better chance the block you want to read is cached, so you’ll get a more comprehensive (read) performance speedup.

Don’t want to invest in new 4 socket servers? You’ll need to take a look at your existing main memory capacity and decide if you want to max that out with this approach. Don’t forget to take any additional vRAM licensing costs into account (if you go over your currently licensed thresholds). If you’ve looked at all this and still want to use Storage Accelerator with Virsto, you’ll get comprehensive performance improvements across both writes and reads.

If you decide against using main memory as a big cache, Virsto by itself still gives you options for comprehensive speedups across both writes and reads. Leverage our tiering to create a fast tier, and put your highest performance disks in that fast tier. You get blazingly fast write performance for all VMs all the time from our logs, and you’ll get great read performance for the shared binaries (that’s the first thing you should look at putting in the tier 0) and any other VMDKs that you’ve decided to lock in there.

But Virsto’s value proposition goes way beyond just performance. Whether you use Storage Accelerator or not, you’ll get huge space savings from our thin provisioned vDisks without giving up any performance, you can fully support features like vMotion and failover (VMware HA), and you can provision high performance, space-efficient storage almost instantly, allowing you to spin up new VMs literally in seconds. And with our support for VMware View, you’ll get access to all this in a manner which preserves all the familiar View workflows in tools like View Composer, View Manager, etc. so there is minimal re-training.

So if you’ve been considering Virsto because we more than double the desktop density your existing storage configuration can support, or you like the fact that you can deploy at your current density using half as many disk spindles, or you like cutting hours off your desktop deployment, re-deployment, or refresh times without any performance impacts, you’re still good to go. With View 5.1, you just now have an additional configuration option for turbo charging your read performance in VDI environments.

What’s wrong with the scorecard?

by Mark Davis, CEO

Posted Thursday, April 26, 2012 in Alternative Solutions 0 comments

Tags: Storage hypervisor

In last week’s post, I thanked the good folks at Ideas International for publishing their storage hypervisor scorecard. Ideas joins other industry analysts such as ESG, EMA, ITCandor, Rainmaker Files, Gartner, Storage Switzerland, IDC, and Jon Toigo in legitimizing the newly recognized storage hypervisor product category.

In this post, I’d like to explain a concern I have about the way some folks are defining the term storage hypervisor.

Categories don’t emerge cleanly

It is never clear in the early days exactly what the defining characteristics of the category are or what to call it.

SUV Do you recall the early days of the product category we now call sport utility vehicles (SUVs)? For years, everybody had a different name and a different definition.

Lots of people try to come up with their own definition. Every vendor that wishes to have their products known as a member of the category has their own spin. Industry pundits of all walks – analysts, reporters, bloggers – weigh in with their own definitions.

We should not expect clarity, especially in the early days of a category.

Is anyone out there old enough to remember the early days of the storage area network? In the 1990s, when the people at Compaq (remember them?) started using “SAN” to define their vision for a new kind of network, to be distinguished from a LAN or a WAN, all heck broke loose for a couple years. Some companies, like Sun Microsystems (remember them?) embraced the term since they had been shipping Fibre Channel storage arrays that supported a whopping two servers. In contrast, lots of industry experts sneered that the SAN category was vile marketecture. Others jumped on the bandwagon and claimed their products belonged in the category, even though many did not build anything that we’d today say fulfilled the criteria for membership in the SAN product category.

So it is, and so shall it be for the Storage Hypervisor

One phenomenon you can bet on happening, especially when there are billions of dollars of market opportunity at stake, is for some people to define the category in a way that is inclusive of old technologies which can’t legitimately fulfill new requirements.

My last post noted that one surefire way to tell that a category definition doesn’t have real meaning is if

“The definition of that category is not clearly distinguishable from any other category.”

Storage Virtualization versus Storage Hypervisor

In their recently published scorecard, an analyst firm defined a storage hypervisor according to the features offered by products in the category. Here are the features they used to score vendors:

Automated tiering
Architectural benefits
Performance monitoring / reporting
Local replication
Remote replication
High availability / disaster recovery
Maximum capacity
Performance scalability
Thin provisioning
Physical provisioning
Virtual provisioning
Migration
Management centralization
Planning capability
Heterogeneous storage pooling
Server hypervisor support / integration

Anything about this list catch your eye?

I found it striking! Maybe it’s because I’ve been involved in storage virtualization since before that category even had a name.

Do you see it too?

If a bunch of storage industry experts were shown this list, and asked what product category was characterized by that set of features, what would they answer?

I’m betting most would say generic storage virtualization.

The list above is in no way distinguishable from a list of criteria for storage virtualization.

And that’s a problem.

There’s no point in a new category defined the same as an old category.

It’s great that the market is beginning to accept the term storage hypervisor. It’s certainly good for Virsto. We started using the term in late 2007. So you can bet we're gratified to have the term gain currency.

But a storage hypervisor isn’t just another name for storage virtualization.

In the next post, let’s plan to review a little storage industry history. There's a cautionary tale to tell.

Thank You Ideas International

by Mark Davis, CEO

Posted Monday, April 16, 2012 in Alternative Solutions 0 comments

Tags: Storage hypervisor

Last week Ideas International posted their inaugural scorecard for storage hypervisors.

As the company that first used the term storage hypervisor, Virsto Software applauds Tony, Angelina, and Joe at Ideas for giving voice to this important topic. Thanks guys!!

We think the topic of storage hypervisors is vital to datacenter and cloud computing.

How vital?

We believe the emergence of the storage hypervisor – with Virsto leading the charge – will transform computing no less than have server hypervisors. That’s right, Virsto’s goal over the next decade is to be as crucial as VMware has been over the past ten years. We have strong feelings about this.

How strong?

We’ve been thinking about and developing a storage hypervisor since 2007. It’s gratifying to have analyst firms like Ideas give the term credibility. We’re so glad to have people join the revolution.

So you might think that I’m really happy about Ideas International’s scorecard.

You’d think wrong

Well, to be honest my feelings are mixed. Of course we’re happy to have Ideas join other analysts like Enterprise Strategy Group, Storage Switzerland, Enterprise Management Associates, Mesabi, among others, in lending their credibility to the storage hypervisor wave. We’ve seen articles about this emerging product category in places like Network Computing, InformationWeek, and TechTarget's SearchVirtualStorage, and I am delighted to see the groundswell of interest from so many.

But overall, I’m not thrilled about the Ideas International scorecard.

Where’s the beef?

My major beef isn’t with Ideas’ ratings of Virsto’s storage hypervisor or of others, although I can quibble with some of the assessments.

My big issue with Ideas’ scorecard is their implicit definition of what a storage hypervisor is.

Words matter. Concepts matter.

I don’t expect that everyone will agree in all respects with my definition of the term storage hypervisor. And I know no analyst can ever fully agree with my definition.

Now that the topic is getting hot, others may catch up with Virsto in their thinking. But for now it’s safe to say that a company of people who’ve been breathing storage hypervisors 7 days a week for more than four years probably have some pretty well thought out ideas on the topic.

The term storage hypervisor has to have some meaning. It needs to be pretty clear whether a product either is or is not in the category.

This is where I have concern about the scorecard published last week.

Two ways to know that a category is too generic

The definition of that category is not clearly distinguishable from another category.
Any vendor who wants to get in on a hot trend can simply rebrand their product as a member of the category.

In a series of posts over the coming weeks I’ll have more to say. A lot more.

We want to get a debate going, so we encourage comments, tweets, articles, blogs, analyses, theses, rants, or musings that link back to our posts here. Again, sincere thanks to our friends at Ideas for prompting what should be a good discussion.

For now, a parting thought

What if, over a hundred years ago, some clever engineers and entrepreneurs invented a new kind of vehicle that had a self-contained mode of propulsion based on internal combustion, electric, or steam power?

What if, back then, they started calling their inventions “automobiles”?

What if, back then, makers of horse-drawn carriages added feed bags to their propulsion systems, and called the things they built automobiles?

And what if no one ever called them on it?

SSD’s Future Will Be About Efficient Usage

by Eric Burgener, VP Product Management

Posted Monday, April 02, 2012 in Technology 0 comments

Tags: performance

There’s no doubt that SSD is here to stay. But it’s not a new technology. Vendors like Texas Memory would sell you a storage device that operated at near memory speeds in the 1980s, but it was so darn expensive that there weren’t many applications where its use could be justified. As SSD prices have come down, there are more applications where its use can be considered.

Given that “performance” is a system-level issue, SSDs are overkill as a way to solve storage performance problems for most environments. It’s like trying to kill a fly with a bazooka. Sure it works, but is there a cheaper way – like maybe a fly swatter – that can achieve the same end result? Those familiar with Amdahl’s law and a little experience benchmarking SSD in virtual environments know that the performance speedup you get from SSD is not related to the native performance of flash but more related to whatever bottleneck you hit next after you’ve removed storage performance as the limiting factor.

What are you really doing when you deploy SSD? If you expect to increase the performance of your system by what the native performance capabilities of flash would suggest, you’re in for a bit of a surprise. What you really do is that you remove storage latency as the performance bottleneck to your system (assuming that is in fact the bottleneck) and now have a new, non-storage bottleneck that you’ll hit way before you get to what the native SSD performance speedup might suggest. Are you familiar with Amdahl’s law? Based on that, the way you deploy is SSD is critical. What you should be thinking about doing is deploying as little of it as possible to remove storage as the performance bottleneck, not just throwing SSD at your “storage performance” problem.

Back in 2008 (I think, or it might have been earlier) Gartner took a look at what was happening to both spinning disk and SSD prices, and predicted that by 2013, both would cost about the same on a $/GB basis. As SSD dropped from thousands of dollars per GB to hundreds and now to tens, we have rightly started thinking about how to use it in more areas. While SSD by itself is really a performance pure play, it does have some other interesting implications given that the single biggest factor you need to take into account when sizing storage configurations of spinning disk has been IOPS (with availability and capacity as secondary considerations in most cases). When you front end spinning disks with some SSD, you end up needing fewer spinning disk spindles to meet a given IOPS requirement , and the cost savings from that can be used to defray at least part of the cost of the SSD.

But the prediction that spinning disk and SSD will be basically equivalent by 2013 is suspect at best. Right now, list prices for enterprise-class SSD (which is more expensive because you need to front end it with NVRAM and software to address wear leveling, random write performance and a few other issues specific to SLC NAND flash) are around $60 - $80/GB at list. If you’ve got a Red Bull-enhanced CFO who can sufficiently beat your storage vendor up, you may be paying street prices in the $25 - $30/GB range. That’s a heckuva lot more than the $1 - $2/GB that that same Red Bull-enhanced CFO is probably paying for your highest performance spinning disk. Sure, SSD is generally at least 10x faster (and sometimes a lot more) than spinning disk, but you’re likely not going to see that in your environment because there’ll be some other roadblock that arises well before that happens.

I attended a meeting of the Minneapolis/St. Paul Computer Measurement Group last month where one of the presenters gave a pretty convincing argument about why enterprise SSD prices are probably not going to go down much going forward. One of his proof points was that on a $/GB basis over the last 18-24 months, SSD has stayed at pretty much the same cost. And that with all the added cost needed to produce enterprise-class SSDs anyway, any future drops in flash prices weren’t going to have too much of an impact on the overall $/GB of those drives. I didn’t verify that for myself, but I did note it.

When SSD is deployed out there, it’s usually deployed as some form of cache. Even if its deployed in a tiered storage environment, its still basically used as a cache. If it’s a write through cache design, you get no write performance speedup, so its purely a read performance play. And even then, only if the blocks you want to read are already sitting in cache. The more SSD capacity you’ve thrown at your cache, the better chance you’ll enjoy read performance improvements. But because SSD is still very expensive relative to spinning disk, you’re being pulled in two directions: to maximize the performance gains you’ll reap you want lots of it, but to keep from blowing your budget you may only be able to afford a little bit of it.

Using SSD in a cache, while it does result in performance improvements, is not a very efficient way to use it since it requires relatively large capacities, and the $/GB metric is where SSD performs worst. By all means, we should be leveraging SSD to improve the overall performance of our systems, but we need to use SSD as efficiently as possible to realize the performance gains that it will actually (not theoretically) produce. In other words, what’s the smallest amount of SSD capacity we need to deploy to move the bottleneck beyond storage performance and to whatever the next roadblock is?

This brings me to my final point, which is about using SSD in virtual computing environments. SSDs are great for read performance, but these environments tend to be much more write intensive than physical computing environments. If you’re looking to leverage SSD in virtual computing environments, caching is not the most efficient way to speed up writes. Logs are a much more efficient way to use SSD to speed up write performance since you typically need at least an order of magnitude less of it, and in many cases even a lot less than that. A log architecture, effectively implemented, will give you write performance speed ups for ALL of your VMs ALL the time. Forget about issues like write through vs write back or whether or not the blocks you care about are cached. Continued high SSD costs (relative to spinning disk) will encourage us to look at better ways to leverage SSD to get what we need out of it.

Looking at ways to address read performance requirements in virtual computing environments more efficiently with SSD is another interesting topic, but I’ll leave that one for another day.

Ready for VDI lift off?

by Gregg Holzrichter, VP Marketing

Posted Tuesday, March 13, 2012 in 0 comments

Tags: Tags, excessive storage spending, Hyper-V, Virsto VDI, virtual desktops, VMware

Virsto VDI Survey: 46 Percent of VDI Projects Stalled Due to Cost and Performance

A survey released by Virsto today finds that virtualized desktop initiative (VDI) projects are of great interest to a majority of medium- and large-enterprise IT organization respondents, however the promise of VDI is being compromised due to cost, performance challenges and end user complaints. Information gathered from the survey reveals a clear disconnect between the growing interest in VDI and the benefits that such projects are realistically able to deliver.

Virsto changes the economics of storage in VDI

Virsto for VDI, first launched in April 2011, is the industry's first VM-centric storage hypervisor that arms desktop architects and virtualization teams with a new approach to deliver storage efficiency and rapid VDI provisioning to remove one of the largest obstacles to VDI adoption: the excessive cost of storage per virtual desktop.

With Virsto, customers can jettison the large costs associated with centralized storage in VDI by getting up to 10x more out of block-based and/or SSD storage capacity, while delivering up to 10x better end user performance, and reducing the time to provision and patch desktops by up to 99%.

This combination of improved performance, increased capacity utilization, and simplified, accelerated desktop storage provisioning acts like a turbo-booster to your stalled VDI project, and delivers cost-effective lift off. In a recent 2,000 seat VDI benchmark with Microsoft for example, Virsto reduced the cost of storage per desktop by over 50%.

Virsto's entrance into the VMware market with support for VMware View and vSphere earlier this year, with support for Citrix XenDesktop on ESX just announced, provides support for all three of the leading VDI front ends, and native integration with vSphere and Hyper-V.

This just might be the year that VDI finally achieves lift off.

What Windows 8 Hyper-V R3 Storage Enhancements Will Mean For Virsto Customers

by Eric Burgener, VP Product Management

Posted Monday, March 05, 2012 in Technology 0 comments

Tags: Hyper-V, Microsoft, virtualization

Last September, Microsoft released some information about what customers could expect to see in Windows 8, and at the very end of February the Windows 8 beta became available. There’s been some interesting activity on the blogosphere about it (see a good blog from Jason Perlow here).

What are the enhancements?

There are a couple of interesting storage enhancements in Hyper-V R3, including VHDX, ODX, built-in deduplication, support for SMB 2.2, and the introduction of ResFS (Resilient File System).

VHDX: Larger Capacity VHDs

In Hyper-V R2, VHDs can be up to 2TB in size, but in Hyper-V R3 with VHDX that has increased to 64TB. While not imposing any limitations themselves, because Virsto vDisks effectively impersonate native VHDs, their size is limited by the native VHD capacity constraints. Virsto welcomes the introduction of VHDX, as this also increases the maximum virtual disk size we support in Hyper-V environments. When Windows 8 ships, Virsto will be supporting vDisks up to 64TB in size. The expanded storage capacity nicely complements increases in CPU and memory on a per VM basis that will allow Hyper-V to host larger, higher performance applications. With Virsto’s unique ability to provide high performance, thin-provisioned, fully cluster-aware storage, Virsto customers will be able to even better support larger, more mission-critical applications with smaller, more cost-effective storage footprints.

ODX: Renewed Invitation to 3rd Party Array Vendors

ODX is somewhat analogous to the VMware APIs for Array Integration (VAAI) concept, allowing certain tasks to be offloaded to storage array hardware to improve performance and/or lower host CPU utilization. But ODX is just an ioctl – the devil’s in the details of implementation, and that falls to storage array vendors and how they implement various features. It’s telling that third party SCSI copy has been around for quite a while, but has never really caught on. Still, we’re rooting for this to take off on Hyper-V because, depending on what the back end storage implementations look like, it could be very interesting to us for two reasons.

First, it may allow us to offload Virsto vLog flushing to an external array, improving the performance of the entire Virsto virtual storage system by enabling higher performance log draining with fewer vSpace spindles. We already generally lower the overall CPU utilization when we’re installed compared to the initial baseline. But this could potentially make it even better.

And second, it’s potentially very interesting for use cases where Virsto customers may want to move data between storage tiers. Today, Virsto supports up to 4 storage tiers. In the next major release of Virsto for Hyper-V, scheduled for 2H12, we will be introducing what we call “dynamic tiering”. This will allow customers to transparently move storage either between tiers or between physical arrays (think impact-less array upgrades) while VMs continue to run, in much the same way that VMware’s Storage vMotion does today. Particularly in cases where you are actually moving data (e.g. array upgrades), the ability to offload that to external arrays bodes well for both performance and host CPU utilization.

De-duplication

Content-based deduplication is a popular topic these days, but it also has the reputation for being a resource hog. Microsoft hasn’t mentioned much about any implementation details of this upcoming dedupe technology (is it file-based? Volume-based? In-line or post process? Where do the cycles come from to perform dedupe operations? etc.) but my guess is that it will be an optional feature that customers can turn on or off. In virtual environments where much of the redundancy is because one or a small number of VMs have been cloned over and over again, Virsto’s approach to data sharing provides a very efficient way to address the issue. And it will be able to co-exist on a single host with any VMs using this new Microsoft feature, giving customers expanded options.

NAS and Hyper-V: Yep, it's in there.

Until now, there really hasn’t been a viable NAS option on Hyper-V. Windows 8 introduces support for SMB, which will allow VHDs to be hosted on shares for the first time. Windows VSS, Microsoft’s snapshot API, has historically operated at the LUN level. In virtual environments, it is much more efficient to perform certain storage operations like snapshots, failover, and replication at the VHD rather than the LUN level. Doing so can save on storage capacity consumption, time, and network bandwidth. With the introduction of SMB, VSS can now operate at the share level as well. If VHDs are hosted on shares, you can potentially do snapshots, failover, and replication with increased granularity and hence, more efficiently.

Virsto has supported the VHD level of granularity for block-based storage for years, providing these increased efficiencies while still allowing customers to enjoy the benefits of block-based storage. That said, we’d still like to see Microsoft make the ability to perform VHD-level storage operations a core part of the VSS infrastructure. That’s an enhancement that would help everyone, and we would readily take advantage of it.

Resilient File System

In ResFS, Microsoft has made reference to a new feature called “storage spaces”. By letting administrators configure storage into separate virtual pools, this enhancement to the storage interface (which until Windows 8 has been through VDS) should enable larger, more scalable, and more easily managed storage configurations. Virsto welcomes this as well, and will reap the benefits of them while continuing to provide added value in the areas of performance, storage capacity consumption, and storage density/cost savings.

CSVs Benefit Too

Finally, I see they’ve also made some CSV enhancements. VSS backup performance for CSVs has been a challenge in the past due to the performance slowdowns often associated with re-directed I/O mode. In Windows 8, it looks like they’ve made some modifications to how copy on write is handled during VSS operations so that updates to a snapshot during the backup process no longer force that VHD into re-directed I/O mode. These changes do not appear to apply to runtime I/O to dynamic and differencing VHDs, just VSS operations, so Virsto will continue to provide significant value when customers require high performance, thin provisioned storage. It looks like they will also improve the scalability of backup operations by supporting concurrent backup of all VMs within a single CSV. Both of these enhancements are welcome additions needed for enterprise deployments. Virsto will still co-exist with CSVs, providing the same high runtime performance, rapid provisioning, and space-efficiency for clustered storage as we have in the past.

The bottom line here is that Windows 8 includes a lot of storage enhancements that Virsto can leverage transparently to make for better configurations, while not challenging any of Virsto’s unique differentiators that Hyper-V customers have found so compelling.

And The Oscar Goes To…

by Eric Burgener, VP Product Management

Posted Wednesday, February 29, 2012 in Technology 0 comments

Tags: virtualization

It’s all fine and dandy that Hollywood awards Oscars to those that work out of the limelight in Tinseltown, but there are a lot of unsung heroes who deserve an Oscar probably more than the guy who won for best original song for the recent Muppets movie. How would you like to hear this: “And the Oscar for best storage administrator performance from a virtual administrator who has had no formal storage training goes to…YOU!”

As virtual computing continues to penetrate mainstream computing, it’s becoming a fact of life, particularly in smaller organizations: virtual administrators are taking on more storage management tasks. It’s not enough that you have to know how to configure multiple VMs on a single host to meet performance requirements, find the perfect balance between security and accessibility for your users, troubleshoot intermittent data corruption problems on your file system, and set up availability clustering to meet stringent recovery time objectives, but you also now have to know how to create relevant RAID configurations, format and mount volumes, and configure and tune tiered storage configurations to meet competing cost and performance requirements. And let me tell you, that’s a LOT harder than writing some Muppet song.

If you’re a larger enterprise, you may still have separate server and storage administration groups, but even in those organizations, everyday tasks like spinning new servers or desktops up and down have storage implications. Having to get storage administrators involved every time that happens increases the time required to respond to user requests and, frankly, pulls storage administrators away from more critical storage tasks that they really DO need to be involved in.

When you spin up a new VM, you generally walk through a series of steps to define a set of characteristics for a virtual disk that you will ultimately mount on that VM. Creating these virtual disks can take time, particularly when you want a high performance one. When you spin down a VM, its storage is released and needs to be reclaimed to a storage pool from which all VM storage is allocated. Vendors like Virsto are simplifying these processes in a way that automatically and almost instantly creates high performance, space-efficient, cluster-aware storage when you’re creating VMs, and transparently handles all the storage reclamation steps, returning that storage to a Virsto-managed pool, when you delete VMs. The intent here is to make common storage provisioning and de-provisioning tasks “no-brainers” so that even a muppet could do it, while giving up nothing in the areas of performance, space-efficiency, cluster-awareness, and speed of provisioning.

To do this right, you want to preserve the native VM provisioning and de-provisioning workflows in popular hypervisors like VMware and Hyper-V, handling this storage stuff pretty transparently in the background. That’s why Virsto’s storage hypervisor is basically just a software plug-in to the major server hypervisor products, preserving the familiar workflows that virtual administrators already know how to use while delivering storage that is optimized for virtual computing environments.

Sure, if you want to change the queue depth on that Emulex or Qlogic HBA in host 7, you still probably want to get the storage administrator involved. But storage hypervisors like Virsto handle the daily storage provisioning and de-provisioning tasks with aplomb, improving your responsiveness while freeing up the storage administrator from routine tasks. And interestingly, with its log-based architecture, Virsto delivers the absolute maximum write performance that your storage is capable of running flat out in 100% sequential mode ALL THE TIME, regardless of what it is. That means that you don’t have to spend much time tuning your storage for performance as your environment grows – that’s handled ONCE when Virsto is first installed - and you always get predictably high performance across operations like adding new VMs, live migration and failover. That frees up even more time on the part of storage administrators, all the while guaranteeing high performance.

OK, those are some pretty bold comments, but we can back them up. If you’d like to understand the technical aspects of why they are true, you can download one of our product white papers off the Virsto website.

Virsto may not ensure that you get your own gold statue, but it will certainly improve your storage performance – by up to 10x - and reduce the amount of storage you need to meet performance requirements – by up to 90% - increasing your server or desktop density per host and lowering overall storage costs. It will lower your CPU utilization by up to 50% and reduce storage latencies by up to 25%. And it will do this ALL AT THE SAME TIME and with your existing storage. Meanwhile, we’ll talk to the academy and see if we can get them to declare a new category for 2013.

An Affair To Remember

by Eric Burgener, VP Product Management

Posted Tuesday, February 14, 2012 in Technology 0 comments

Tags: excessive storage spending, Hyper-V, performance, VMware

It’s Tuesday, February 14, Valentine’s Day, and love is in the air. When you think about the “great couples of the ages” – Marc Antony and Cleopatra, Romeo and Juliet, Bogie and Bacall, etc – would you ever place your relationship with your hypervisor in that group? Besides being an economic stimulus package for florists, Valentine’s Day does give us the opportunity at least once a year to think more deeply about our relationships.

Maybe you don’t think you have a “relationship” with your hypervisor. But think about this – you probably spend at least 4 hours a day, and maybe as many as 12, working closely with it, some days you curse it, and some days you thank your lucky stars for it, sometimes even after a full day you still contact it from afar (remote login) because 12 hours just wasn’t enough, and you probably can’t go on vacation without staying in touch with it. You’ve may even have had to leave a romantic dinner or two with a real person just to touch base with it when it got out of sorts. Like it or not, you two have a relationship.

Well, today is a day to reflect on the good in that relationship. Your hypervisor has probably made your administrative life a lot easier, letting you manage a lot of tasks centrally that you used to have to travel to address. You can be much more responsive to the needs of the business, something which bodes well for your relationship with your boss. But in administration just as in life, there is no free lunch. And the cost of the joy your hypervisor brings you has often been paid in storage dollars.

Real relationships require a whole host of ancillary services that help you keep that spark alive with your significant other – Hallmark, FTD, Godiva, high end restaurants, Victoria’s Secret, monster truck shows (hey, that’s why there’s vanilla AND chocolate!), cruises, etc. – and now, there’s one to help keep you and your hypervisor on the road to eternal bliss: Virsto.

What could be more romantic than giving you up to 10x more performance out of your existing storage without buying any extra hardware? Provisioning high performance storage in seconds? You like thin? We’ll give you thin provisioning AND high performance for your storage. We package everything up into a storage object that looks exactly like your native hypervisor’s virtual disk, so it fits right into those familiar and comfortable management routines you already share with your hypervisor but gives you new capabilities to help keep your relationship fresh. And we’ll give you all this using your existing storage, whatever it is. This is a romantic interlude you can share with your favorite hypervisor (or two, since we support BOTH vSphere and Hyper-V) without breaking the bank. It’s the magic of pure software, and it puts the Shane company’s options to shame.

Virsto will address your storage performance issues, regardless of whether you’ve got a virtual server or a virtual desktop infrastructure environment, and we’ll lower the cost to keep your relationship going – you’ll need a lot less storage to meet your performance requirements. It’s like buying a single rose but getting the full blown dozen long stems, baby’s breath and all.

OK, so maybe adding Virsto to your virtual environment won’t inspire you to write the sequel to Omar Khayyam’s Rubaiyat, but it will help you to work through that nagging hypervisor relationship issue so many of us experience – “but honey, your storage is just so darn EXPENSIVE!” – without having to bring in a counselor. And who knows, it may get you to see Khayyam’s most famous quatrain in a new light (with appropriate nods to Fitzgerald):

“A Book of Verses underneath the Bough,

A Jug of Wine, a Loaf of Bread – and Thou

Beside me singing in the Wilderness –

Oh, Wilderness were Paradise enow!”

Lightning, thunder and turbines: Oh my

by Mark Davis, CEO

Posted Thursday, February 09, 2012 in Alternative Solutions 0 comments

Tags: excessive storage spending, performance, virtualization

All of the talk of lightning, thunder and turbines this week has evoked a stark image of a jet plane hitting stormy weather. It’s an apt metaphor for what IT is facing with storage in virtualized environments. CIOs we speak with confirm that storage budgets have been hit with a force 5 hurricane due to the widespread adoption of server virtualization. One thing’s for sure: neither PCIe SSD cards nor SSD storage appliances will provide shelter for buffeted storage budgets.

EMC’s VFCache and the more established Fusion IO certainly can support more I/O, and that’s a good thing for virtual machine workloads. These approaches validate what we’ve been saying for a long time at Virsto: that getting closer to the source of the problem (virtualized workloads in the server) can yield superior results than overprovisioning hardware (solid state and otherwise) in an external disk array. Clearly, this is where storage architectures need to move.

One could deal with VM performance challenges only with specialized physical cards that install in the server, physical appliances stuffed with SSD to replace the array, or caching approaches that rely heavily on SSD. Nothing wrong with cool hardware, and at Virsto we’re huge fans of SSD. But data center architects have to be careful not to perpetuate a very nasty habit: throwing excessive hardware at problems to overcome weaknesses in software architecture. Over the years, we’ve gotten far too comfortable with overprovisioning storage arrays. Please, for the good of all humankind, let’s not do the same things with server-side storage, ok?

A more holistic approach already exists for IT to bring server and storage virtualization into equilibrium: It’s called a storage hypervisor.

We at Virsto happen to build what many think is the best one in the business.

Yeah, performance matters. But there’s more to it than only performance.

Yes, one of the fundamental challenges caused by server virtualization is highly random I/O, and yes, flash based SSDs can be good tools to solve one of the VM I/O problems. (Although, as people in the know will tell you, flash is not inherently very good at small block random writes, which is what server hypervisors tend to generate gobs of, especially in use cases like VDI.) However, a brute force approach to solving just the performance aspect of the VM I/O conundrum can be just another form of overprovisioning (that is, overspending on) storage hardware.

The Virsto approach to the broad set of VM I/O problems is holistic. Yes, we deal with performance issues. Just last week a customer in Europe bought our software because we made their hardware do I/O 44X faster. And no, that is not a typo. Adding software, the Virsto storage hypervisor, to this customer’s existing hardware made it go 44 times, or 4400 percent, faster. When we say we can relieve you of the burden of over-spending on storage hardware – on both the server and array ends of the cable – we really mean it.

We made their hardware do I/O 44X faster.

But there is more to the VM (and cloud) I/O problem than just random I/O performance. That’s why Virsto has a more holistic view.

There is also the issue of massively wasted terabytes. I hate to be the one to tell you (let alone your CFO), but perhaps 80% of the physical terabytes you buy for VM storage are not really needed. And let’s not forget the billions of dollars in operating expense wasted by slow and complex VM provisioning. I could go on. There are plenty of other VM I/O problems that SSD, whether server-side or array-side, aren’t built to fix.

About, oh, ten years ago, we all could have kept dealing with our server problems by throwing expensive and proprietary hardware at them. But some pretty clever folks in Palo Alto, Cambridge, Redmond, and around the globe came up with a better idea. It was called a server hypervisor, and the concept changed the world in a fundamental way.

Now comes a team of people in Sunnyvale and Melbourne who have spent a couple years building a fundamentally different and architecturally elegant way to deal with the totality of VM I/O problems: the Virsto storage hypervisor. It boosts performance, with or without SSD, sure. But there’s a whole lot more.

SSD alone will not solve all the fundamental challenge of storage in virtualized environments. The emergence of the storage hypervisor will bring the economics of storage back into line. Just in time for things to get truly cloudy and stormy.

Just What Is A Storage Hypervisor?

by Eric Burgener, VP Product Management

Posted Thursday, December 15, 2011 in Technology 0 comments

Tags: excessive storage spending, Hyper-V, virtualization, VMware

There’s a new segment emerging in the industry around storage hypervisors. This is a simple idea that makes a lot of sense. We all understand the benefits that server hypervisor technology brought to servers in terms of utilization, flexibility and cost savings. The promise of a storage hypervisor is to do for storage what server hypervisor technology did for servers, providing the same utilization, flexibility, and cost savings benefits.

Keeping with the analogy, a storage hypervisor must meaningfully improve the utilization of existing storage resources to drive lower costs and second, it must provide the level of virtualization necessary to offer unfettered management flexibility.

Let’s face it: we’ve been overprovisioning storage for years. The wastage this caused has driven the development of some interesting technologies like storage capacity optimization (the most well-known example of this technology is de-duplication) and thin provisioning, among others. Most enterprises didn’t really have a good handle on just how badly they were underutilizing storage resources along the lines of performance and capacity consumption. A “storage resource management” push about a decade ago attempted to help enterprises quantify this but not surprisingly, storage administrators weren’t too keen on making it easy for others to see just how underutilized the storage was.

Fast forward to virtual computing. As badly as storage was underutilized on physical servers, its much worse on virtual servers. The I/O patterns in virtual computing environments are much more random than they ever were on most physical servers, and spinning disks don’t handle randomness very well. The more randomness in the I/O pattern, the fewer IOPS the drive produces. So you end up with drives that are producing maybe only 65% of the IOPS of which they are theoretically capable, which means you have to buy more drives to meet your performance requirements. If you could just use the drives you have more efficiently, you’d need a lot fewer of them.

Storage capacity isn’t used very efficiently in virtual computing environments either. If you’re running in production, you’re very likely to be using thickly provisioned virtual disks to ensure you get the performance you need. Provisioning this storage is not very efficient - it takes a long time - and it wastes storage capacity by pre-allocating a lot of storage that you may not use for months or never actually use. Thin provisioning by itself is not the answer here, because it often doesn’t provide the performance you need. What is needed is high performance thin provisioned storage.

If you haven’t seen the impact of this conundrum already, as your environment grows you will. You’ll find you need 30% - 50% more storage hardware to meet your performance requirements in virtual computing environments.

And what about management? If you’re using block-based storage, you are managing storage at the LUN level. But as a hypervisor administrator, what you’d really like to do is have the storage administrator put your configuration together up front, and then thereafter just manage at the virtual disk level (VMDKs on vSphere, VHDs on Hyper-V) so you could handle storage provisioning and de-provisioning yourself. When you perform storage operations like snapshots, live migrations, failover, or replication, you want to perform those operations on a select group of virtual disks, not on every virtual disk that happens to be on the same storage LUN as the ones you care about. And you want to take advantage of rapidly provisioned, high performance, space-efficient storage when you do that. This is where storage virtualization technology comes into play.

Before we sum up, I’d like to comment on SSDs. SSDs have been the industry’s response to virtual computing’s storage issues, and prices on them are definitely coming down. It’s clear that SSD has a place in virtual computing going forward, but just throwing SSD willy-nilly at storage performance issues is somewhat analogous to trying to improve physical server environments by just buying a faster CPU. You already had CPU cycles that you weren’t using. The real innovation with server hypervisors was to change server architectures to use existing CPU cycles more efficiently. Then, if you wanted to buy a faster CPU, you’d get even more out of that too.

Think about how this applies to storage. I have drives that theoretically could deliver 180 IOPS, but each is only delivering 135 IOPS on average. I’m allocating storage that I’m not using so I could actually “run out” of storage while I have tens of terabytes of allocated but unused storage sitting in my array. I can spin a VM up very quickly if the storage is already provisioned, but provisioning high performance storage takes a long time – it’s not very efficient. And if I’m managing at the storage LUN level, then I’m not performing storage operations like snapshots, live migration, failover, or replication very efficiently in terms of time, capacity consumption, or bandwidth utilization. A storage hypervisor changes things so that these pre-existing resources are operating much more efficiently. Then, if you want to add SSD on top of that, you’ll be getting a lot more out of that as well.

So to sum up, there are 4 areas that a storage hypervisor needs to address to improve storage resource utilization: performance, capacity consumption, provisioning, and management. The bar on server hypervisors has been set, and the industry is adopting that technology by storm. It’s time for a storage hypervisor, and server hypervisor technology provides a great analogy to help define exactly what a storage hypervisor does.

Re-Examining NAS vs SAN in Virtual Computing Environments

by Eric Burgener, VP Product Management

Posted Wednesday, October 26, 2011 in Technology 0 comments

Tags: excessive storage spending, VMware

I’ve been spending a lot of time at trade shows and with customers in the last couple of months talking about our upcoming vSphere product scheduled to ship by the end of calendar Q411. From these conversations, a good number are considering NAS in their vSphere environments. When asked about why they find NAS attractive from a storage point of view, I’ve heard the following:

NAS is easier to manage than block-based storage
NAS is by definition shared storage so I can support desired live migration and high availability capabilities
NAS gives me thin provisioning to save on storage capacity without me having to consciously configure it, but this often evolves into a conversation about the ability to quickly provision high performance storage for new VMs
I can use array-based snapshots, which offloads the host and can offer performance and scalability advantages over hypervisor-based snapshots

As I review the broad set of use cases in virtual computing, there’s no doubt that all of these features are desirable for various reasons. But I’ve noticed an undercurrent in these conversations – an assumption that you can’t get these capabilities together - especially the ease of use one - with block-based storage. Let’s look at a conversation I’ve had a lot recently.

First, let’s examine ease of use. Administrators say that NAS (i.e. file) is easier to manage than block-based storage. My end users care about files, not LUNs, and NAS lets me manage things at the file-level instead of at the LUN level. It makes management VM-centric instead of storage-centric (after all, VMs are what I really care about) and has implications for live migrations and snapshots – I can save time, bandwidth, and space by just working with relevant VMDKs instead of entire LUNs that inevitably have multiple VMDKs on them, some of which I care about for a particular storage operation and some of which I don’t. Also, NAS manages all the space reclamation issues for me when I’m deleting VMs, so that’s just one less thing I have to do if I’m spinning VMs up and down a lot (or even only sometimes).

Second, shared storage. HA is a requirement in many production environments. Shared storage is a pre-requisite for moving VMs around between hosts, regardless of whether you’re doing it for HA or for load balancing reasons. FC SANs give me a shared storage capability, but they are block-based, and if I’m interested in making storage easier to manage that could be a knock against them.

Third, thin provisioning. Thin provisioning allocates storage only as I need it, which in many virtual computing environments can save up to 90% on storage capacity consumption relative to the older approach of allocating all storage required for a particular VMDK up front. And I can provision thin provisioned devices very quickly when I’m spinning up new VMs – much more quickly than fully allocated “thick” virtual disks. But there’s that pesky issue of performance. There’s no doubt that thin provisioned devices perform more slowly – in some cases as much as 90% slower – than thick disks, but they may perform fast enough that “performance” isn’t an issue in your environment. What would be really interesting, though, is if you could provide a thin provisioned device that performed at the level of fully allocated devices right up front. This would directly impact the VM density your storage configuration could support, and if this lets you deploy, let’s say, twice as many VMs per host while still meeting performance requirements, now that starts to become a lot more interesting.

Fourth, snapshots. When people talk about NAS giving them access to great snapshot technology, they’re generally making a veiled reference to NetApp. VMware provides a VMware API for Array Integration (VAAI) for both block and NAS-based devices, but until recently VAAI NAS was not accessible to any vendor other than NetApp. vSphere 5 promises to open that up to other vendors. VAAI NAS allowed snapshot/clone creation and usage to be offloaded to a NetApp array, which allowed VMs to access a more performant, more scalable snapshot technology than was available in the ESX hypervisor.

There’s a lot to like here. But block-based FC storage has some nice advantages as well, particularly in the areas of performance and security. A lot of folks already have FC SANs, and might just plan on expanding them as they roll out their virtual computing environments. And frankly, buying NetApp NAS with all the trimmings for snapshots and performance isn’t cheap. Can’t somebody give us an option that gives us the performance of FC with the generally recognized advantages of NAS?

For me, this is probably the core technical differentiator that Virsto brings to the table. We run on block-based storage, and if its shared storage then we fully support capabilities like failover and vMotion. Our logging architecture (implemented at the storage layer) ensures that you’ll get performance that exceeds that of fully provisioned “thick” virtual disks with storage objects (which we call Virsto vDisks) that are thin provisioned so they can be provisioned extremely rapidly - we're talking literally thousands of clones in seconds for tasks like, say, desktop provisioning. You’ll get this same performance and rapid provisioning on any snapshot and clones you create, regardless of how many of them there are. And we support a lot – you can realistically create and use tens of thousands of snapshots/clones without any performance degradation. And finally, we give you the ability to manage storage operations at the level of granularity that you need for maximum efficiency – the VMDK level.

We’re working on getting this story out, but its an evangelical task. Most folks I talk to have pre-conceived notions about what SAN and NAS give you, that SAN is hard to manage, that thin provisioned devices perform slowly, that you have to pay a lot to get scalable snapshot technology. Are you operating under some of these same assumptions? They were true in the past, but the times… they are a-changin’.

Virsto enters VMware market, unleashes storage hypervisor

by Mark Davis, CEO

Posted Monday, August 29, 2011 in Events, Technology 0 comments

Tags: Virsto VDI, VMware

Today Virsto officially entered the VMware market with the launch of Virsto for VDI, vSphere Edition. With our announcement, Virsto continues to deliver on the promise of the industry’s first storage hypervisor, with a software-based approach that simplifies the management of storage in virtualized environments and lets customers get much more out of their storage hardware investments. I believe our approach disrupts traditional storage in the same way VMware’s hypervisor transformed the server market. VMware didn’t enter the server market until 2001, but it’s hard for most of us to remember data center life before the impact of virtual servers.

The industry is recognizing that the storage side of virtual computing needs to catch up with the server side evolution that has occurred over the last ten years. Specifically, we need a hypervisor for storage just like we have a hypervisor for the server computing infrastructure. And that’s what we at Virsto set out to do in 2007. We believe that doing this well--and in a way that’s relevant to the industry--requires supporting all of the major virtualization use cases: server consolidation, test and development, virtual desktops and cloud computing. It also requires the ability to abstract the logical from the physical hardware, so that any storage hardware can be utilized effectively on any virtualization platform.

At least 90 percent of the customers I talk with have more than one hypervisor in place. The reality and the future is that of the multi-hypervisor data center. With the announcement of our first VMware product today, Virsto is demonstrating that a multi-hypervisor storage solution can provide customers with a consistently high level of storage functionality that is totally optimized for virtual machines, in a way that lets customers derive value from any hypervisor they are running. Until today, no product has ever been able to do this.

Our entrance into the VMware marketplace is an important step in our corporate strategy — but by no means the last step. Over the past four years, everything we’ve done at Virsto has been working toward this larger vision of supporting all the major hypervisors, which today’s announcement denotes.

We’ll be showing the latest in booth #260 at VMworld this week — stop by and see the future of storage in action.

Clarifying How Virsto Plays with Hyper-V R2 Cluster Shared Volumes

by Eric Burgener, VP Product Management

Posted Monday, August 22, 2011 in Technology 0 comments

Tags: Hyper-V, Microsoft, performance, Virsto One, Virsto VDI, Virsto VSI

On Hyper-V, a good number of our customers are using failover clustering. A few questions have been surfacing about how Virsto can best be used in a failover cluster environment, so I thought I’d make a few comments about that.

First, let me state that Virsto vDisks co-exist with CSVs in every Windows Server Failover Cluster environment we have in our customer base. A CSV is a volume that is available to directly read from and write to by all nodes in a failover cluster, and it’s the best place to put the VM configuration data in a failover cluster. Virsto’s user manuals actually recommend that customers configure a CSV for this data in every Virsto Server Pool that is also a failover cluster. So if you’re using Microsoft failover clusters with Virsto, you will have at least one CSV in every configuration.

Where we recommend that Virsto vDisks be used in a failover cluster is for the VM image and user data. This is where some of Virsto’s unique features have real value – the ability to place this data on very high performance, thin provisioned Virsto vDisks that support very scalable snapshot/clone technology makes Hyper-V perform even better during runtime operations as well as during VSS/DPM backups (which by the way, Virsto fully supports). Virsto supports very granular operations at the VM level – not the LUN level – even though its running on top of block-based storage. That means you get the flexibility you need to Live Migrate individual VMs where you want them without having to worry about how you’ve laid out your underlying LUNs.

By the way, Virsto just received Windows Server 2008 Certification, just to underline our ability to fit into Hyper-V environments. If the logo is not up on our website yet, it should be up shortly.

Let’s just make sure everyone knows what a vDisk is. Virsto CTO Alex Miroshnichenko has already covered the detailed implementation of vDisks in a Hyper-V environment, but it wouldn’t hurt to review that again. A Virsto vDisk looks to Hyper-V exactly like a Hyper-V fixed VHD, but it is thin provisioned and runs at the same performance as a pass through disk regardless of whether it has any snapshots or clones attached to it. To understand why this is true, you can read up on our virtual storage architecture on our website. This Virsto “virtual VHD” is only a namespace object - just an entry in a virtual directory. The Virsto filter driver, which is what you install in the parent partition of each Hyper-V Host, maintains the linkage between a Virsto VHD namespace object and the corresponding vDisk that contains the actual data. By attaching or mounting a vDisk to a Hyper-V VM, this linkage is established. Once that’s done, all the I/O requests flow directly between the VMs and the appropriate vDisk.

To sum up, it’s the combination of Hyper-V, failover clusters, CSVs, and Virsto vDisks that together create a compelling hypervisor platform for either virtual server or virtual desktop deployments that trounces the hypervisor competition. Virsto enhances this Hyper-V configuration so that, with a given storage configuration, it will support a virtual machine density that is at least 2x higher than native ESX while it at the same time outperforms it on an IOPS/spindle basis. And it provisions faster, using our scalable snapshot/clone technology, than a high end, enterprise-class storage array, but using any heterogeneous block-based storage. This all translates to significantly lower cost Hyper-V solutions that give customers a real reason to look at vSphere alternatives.

What Kind of ROI Do You Need To Bring A Technology In-House?

by Eric Burgener, VP Product Management

Posted Tuesday, July 12, 2011 in Technology 0 comments

Tags: dedupe, Hyper-V, Virsto VDI, Virsto VSI, virtualization, VMware

We’re probably all generally aware that deduplication technology has entered the market pretty successfully. A recent survey done by TechTarget (May 2011) indicated that slightly more than 60% of respondents are using dedupe in backup, and 57% of the respondents who are not using dedupe expect to implement it within 2 years. There are lots of vendors playing in this space, the first multi-billion dollar acquisition (Data Domain) has occurred, and storage capacity optimization (SCO) technologies are also starting to worm their way into primary storage environments, with at least one telling acquisition already behind us there (IBM’s acquisition of Storwize). The headliner benefit of dedupe? Storage capacity savings. Survey responses indicated that, on average, survey respondents saved 43% on their disk capacity consumption for backup. That’s almost a no-brainer.

Whether you deploy dedupe technology through appliances, backup software, or virtual tape libraries, it’s not cheap. But that doesn’t really matter, because it pays. If before dedupe you were maintaining a 50TB secondary storage (i.e. backup) capacity, that would have on average have dropped to around 28TB. That 22TB savings, if we assume a $/GB price for midrange, raw disk capacity at around $15/GB (add up the cost of the cabinet, trays, I/O modules and disks and divide by usable raw capacity), is over $300K. Let’s say you spent $150K on your dedupe infrastructure – that still leaves you with $150K savings pretty much right off the bat, plus you get the added advantages of faster backups, probably faster restores (because you’re transferring the smaller, deduped amount across the network), energy savings, and less secondary storage to manage. And you get to leverage this type of savings moving forward as your primary storage environment grows to help keep backup issues (time, capacity, costs) in check.

If payback like that meets your ROI requirements, what about if you could enjoy additional storage capacity savings of that magnitude for $2,800 per host in your virtual environments? Check this out:

The average Virsto customer uses 74% less storage capacity than they did before they deployed us (without counting any backup capacity savings) – on the example above, that would have resulted in over $550K of savings, $225K MORE than dedupe; $330K or $555K savings - which would you choose? how about both?
Unlike dedupe where there is a question of whether or not it may impact performance, Virsto uses a unique logging architecture to sequentialize hypervisor I/O, significantly improving performance (specifically IOPS/spindle)
By moving your storage solution (whatever it is) to a much higher utilization rate, we increase the virtual machine density that any given storage configuration supports by at least 2x

These numbers are based on joint testing Virsto did with Microsoft at their Enterprise Engineering Center in Redmond, WA in the last 6 months and validated by an independent third party (the Veritest arm of Lionbridge). If you’d like more info on these tests, please contact us and we can share the performance white paper on them.

Using Virsto isn’t an either/or proposition when it comes to dedupe – we can be deployed with dedupe. If we’re deployed with it, we reduce the amount of primary storage you have to back up, which reduces your secondary storage as well. That’s before any storage capacity optimization benefits you may get from the dedupe technology. Much of that 74% savings above comes from our very efficient thin provisioning implementation, so it won’t conflict with the benefits dedupe can provide. Or deploy us without it - either way, you get substantial savings even if you just look at capacity, forget about how much you save from the reduced spindle count that comes from at least tripling the IOPS that you'd get out of your current thin provisioned virtual disks.

But, you might say, I'm not using thin provisioned virtual disks. Why not? Their poor performance. Wait, you at least triple performance of the Hyper-V VHDs? OK, now I'm starting to get it. Thin provision my VHDs to save on capacity, but still get the performance I need for my production environment...

You may have already decided what kinds of capacity savings it takes to bring storage capacity optimization technology into your shop, and Virsto will likely beat that ROI by a handy amount. Plus it will speed up the storage you already have (or any new storage you want to buy) and give you scalable, high performance snapshot technology that outperforms what you’d get on high end enterprise-class arrays using any block-based storage on the back end. But don't forget - this type of savings could be additive to anything you get with dedupe.

Now that really is a no-brainer.

Go after the low hanging fruit first

by Eric Burgener, VP Product Management

Posted Wednesday, June 29, 2011 in Technology 0 comments

Tags: excessive storage spending, Hyper-V, Virsto VDI, Virsto VSI, virtualization, VMware, Xen

In the May issue of Storage Magazine, editor Rich Castagna published the results of a twice yearly survey TechTarget does concerning planned storage spending. This time around, there were 833 respondents with an average company size of $1.4B. According to the survey, disk capacity growth is starting to gain momentum again, which is a change from the last 2 surveys where it had stalled. Those surveyed this time around said they’d add an average of 43TB in 2011. Going the next level down, larger enterprises expect to add 94TB, midsized enterprises are at 43TB, and even small companies are looking at adding an additional 22TB of capacity.

That’s a lot of storage. Is it all necessary?

Disk is cheap, and just adding disk to deal with storage growth is probably the easiest short term way to address it. All the other approaches tend to require a lot more thought and probably some new processes – basically, changes that most prefer just not to deal with. The truth sometimes hurts, but the reality of this is that storage is expected to continue to grow at these rates for at least the next 5 years (according to IDC) and just throwing hardware at the problem at some point won’t be enough. Ultimately the administrators who manage storage are going to have to implement new processes that, while possibly already available today, are not being widely used. The question for you is: when does that happen for me?

You’re probably already thinking about how to address this problem, even if you’re not actually doing anything other than just buying more storage capacity in the near term. Your storage, energy, and administrative budgets are increasing along with your storage purchases. There are a lot of technology options – storage capacity optimization, storage tiering, archiving, data classification, storage quotas, storage virtualization, etc. - but what's the easiest thing to do?

I’d suggest you address the low hanging fruit first. If there are things you can implement that are simple, don’t require exotic hardware purchases, are transparent to end users, and don’t increase your administrative burden but have a real impact on how much storage you need by increasing your ability to utilize what you already own much more efficiently, that might be of interest.

What if you could install a software component in your hypervisor that would triple IOPS/spindle and decrease storage capacity consumption by over 70%? What if this component worked not only on Hyper-V, but also on ESX and XenServer, and allowed for storage to be moved between hypervisors without having to import any data? With it, you speed up your existing storage while at the same leveraging thin provisioning, providing very scalable snapshot technology (we’re talking tens of thousands of snapshots), fully supporting VM migration (Live Migration, vMotion, XenMotion), and getting a few other goodies you’ll find out about if you delve more deeply into what Virsto Software offers.

If you’re a large enterprise, instead of buying 94TB this year, you might only have to buy 28TB, and you’ll also spend less on rack space, energy, backup, and administration. If you’re paying $10/GB for your storage (well within the range of enterprise-class storage), you’d save well over $600K on storage this year. And other than installing and initially configuring it, you wouldn’t have to do anything differently from what you’re doing now. It’s simple, doesn’t require any hardware purchases, is fully transparent to applications and end users, and actually reduces your administrative burden while improving storage performance and storage utilization.

If you’re using virtual computing in your infrastructure and this sounds interesting to you, I’d invite you to check us out at www.virsto.com.

Fresh financing, and an acquisition

by Mark Davis, CEO

Posted Thursday, June 09, 2011 in Software Business 1 comment

Tags: startups

Yesterday Virsto announced a $12 million series B round of venture capital financing. We also announced our first acquisition.

Big milestones for a young company.

What this financing means

Series B financing The investor appetite for Virsto was high in this fundraising process. Lots of interest from lots of firms, ample willingness from VCs to write significantly large checks, and on terms that are of course good for the investors but are also favorable to Virsto.

Completing this strong of a fundraising validates

the multi billion dollar market opportunity Virsto is pursuing,
the technology Virsto has developed that is at least 2 years ahead of anything in the market,
the customer and partner momentum Virsto has built,
the team Virsto has assembled,
and Virsto's strategy for growing far beyond where we are today.

I couldn't agree more with the Sand Hill Road venture capitalists on all these fronts. As with the previous $8.25 million raised, we'll use this capital to broaden our solutions, enter new markets, and build partnerships.

What the acquisition of EvoStor means

After we started Virsto, we heard of other small teams starting up and trying to build solutions to some of the same virtualization storage problems. Each startup had its own angle, different team backgrounds, and a few actually had some financial backing to fuel their pursuit of the opportunity.

One such company in Virsto's market space was EvoStor. They were of roughly the same vintage as Virsto.

EvoStor took a very different approach from Virsto. While Virsto chose Microsoft Hyper-V for our first entry point in the market, EvoStor focused on VMware. Virsto is 100% software, EvoStor embedded software in appliances. Virsto centered R&D in Silicon Valley, EvoStor in Melbourne, Australia. Despite those differences, there was also a lot of commonality in our belief in the need for a new storage architecture for virtualization and cloud.

We didn't cross paths that much, but we at Virsto were always aware of EvoStor. Customers, partners, and VMware executives all raved about the slick, incredibly powerful and easy to use integration EvoStor built into the VMware platform.

When we saw the opportunity to acquire the EvoStor software engineering team and the technology they built, I knew it was worth trying to find a way to combine forces.

When we got to know each other, we found the expected (and obvious) short term synergies between the two companies' technologies. We were surprised, however, and pleasantly so, to find in EvoStor expertise, vision, and software code that will impact the Virsto longer term roadmap in significant ways.

A half dozen of the key EvoStor engineers have been working on Virsto projects since March 1, so a lot of progress has already been made. We can't wait to show you a demo! (If you want to sign a confidentality agreement with us, we have some things to show you now.)

EvoStor didn't quite get to the point of launching their products before becoming part of Virsto, so the products they were developing will never see the light of day. However, you can bet your bottom dollar (whether US or Australian) that their technology, vision, expertise, and passion will show up in Virsto products very soon.

Hyper-V enlightenment from a beer commercial?

by Eric Burgener, VP Product Management

Posted Thursday, June 09, 2011 in Technology 0 comments

Tags: excessive storage spending, Hyper-V, performance, Virsto One, Virsto VDI, Virsto VSI

Remember the old Miller Lite commercial from the 1980s that had two distinct sets of committed adherents? One set of guys earnestly opined that they preferred the beer because it had “great taste” (debatable at best if you’ve ever actually tasted it), while the other was just as passionate about the fact that it was “less filling” (probably true). The ladies even got into the mix on this one. I can just imagine Hyper-V systems administrators, just as emotionally charged, having heated debates about which is better when it comes to virtual hard drives (VHDs) – high performance or thin provisioned.

Now this argument would be applicable if you’re just looking at the native Hyper-V VHD options. Of course, there’s another option. Instead of drinking a lite beer at around 60 cents a can, you could choose a bottle of 1994 Screaming Eagle Cabernet Sauvignon at $2,000 a bottle. They’re both designed to address the same problem – thirst – but they each get there a little differently. In storage terms, this is the difference between the cost of using a native Hyper-V VHD option or buying a high end enterprise class array where your “by-the-drink” price (i.e. $/GB) is going to be a lot higher. Sure, if I buy a high end NetApp or EMC array, I’ll get both high performance and thin provisioning, but isn’t there another way to do the same thing that is a little more affordable?

Actually, there is, and that’s where Virsto comes in. Install our software in the parent partition of a Hyper-V node, and you’ll get VHDs that are both high performance and thin provisioned at the same time. But it doesn’t stop there. When you use a Virsto VHD (which our marketing guy cleverly calls a vDisk), you also get very scalable, high performance snapshot/clone technology and cluster-awareness for full support of Hyper-V Live Migration that you can use with any block-based storage. This is like getting a full layout of prime rib, cooked rare, with a baked potato and fresh seasonal vegetables, to go with that bottle of cabernet.

Oh, and did I mention that your cluster-aware vDisks, which are managed exactly like native Hyper-V VHDs, will support very high performance snapshot backups using Windows VSS and Microsoft DPM? Add a Caesar salad and white glove table service to that bottle of cab.

So if you’ve been agonizing over which group you’re in (great taste/high performance or less filling/thin provisioned) when it comes to choosing storage options for your Hyper-V environment, you now know how to get both without breaking the bank. If you'd like Screaming Eagle cab "taste" at lite beer prices, you might want to check us out. But while our customers are passionate about what we provide for them, I don’t expect to see mud wrestling youtube videos as sys admins hash out their Hyper-V storage preferences anytime soon!

The Business Argument for Higher Storage Utilization

by Eric Burgener, VP Product Management

Posted Wednesday, June 01, 2011 in Technology 0 comments

Tags: excessive storage spending, Hyper-V, performance, Virsto VDI, Virsto VSI, virtualization, VMware, Xen

Here’s a scenario for you. You’re put in charge of manufacturing operations for your company, which is selling the widgets you make as fast as you can make them. You discover that you are running your manufacturing operations at 35% of capacity. What would you do?

First, you’d probably ask why. Then you’d probably compare the cost of running them at a much higher utilization rate with the additional revenue from sales. Chances are that you would almost immediately move to a plan that more fully utilizes your existing production capacity.

In introducing your company to server virtualization technology, you may have made a somewhat analogous argument to executive management. If you were running your physical server infrastructure at roughly 15-20% of its capacity (which, by the way, is pretty much on target with what most companies were doing), that left a lot of capacity that you had already paid for unused. As you grow and need more computing capacity, you could add more servers, using 15-20% of each new server, or you could look at an approach which increases utilization of existing servers first to accommodate growth and help keep costs down. In moving to server virtualization technology, you could promise your company that you would easily triple server capacity utilization while at the same gain some other benefits in the areas of flexibility to adapt to an evolving business environment, easier management, and lowered energy costs. Bottom line: this change will allow us to get a lot more out of the IT infrastructure that we pay for.

When you made the move to virtual servers, chances are you stayed with the same old storage architecture. Sure, maybe you moved to a networked storage environment so you could more fully leverage what server virtualization had to offer, but the underlying storage architecture was the same. Why? The same capacity utilization issue exists with legacy storage architectures that existed with physical servers and with the manufacturing capacity example above. If you’re like most companies that have moved to virtual servers, you’re running your storage about a third as efficiently as it could be run.

There are two key facets to this definition of “efficiency”: performance and capacity. IOPS/spindle is a relevant measure on the performance side for most application environments, while the amount of allocated storage capacity relative to the amount that is actually used is the key one on the capacity side. Because of the extremely random, very write-intensive I/O patterns generated in most virtual computing environments, the IOPS/spindle you get from spinning disk in these environments is about a third of what you would get from that same disk running in the old client/server model with one application running on a dedicated physical server. And because you’re already at a disadvantage in terms of IOPS/spindle, you may have decided against using the native thin provisioning technology included with the major hypervisors (ESX, Hyper-V, XenServer) to address the “efficiency of space utilization” issue – you can’t afford the performance hit.

If you apply the same logic used above for the manufacturing and initial move to virtual servers, you’d start looking for a way to increase the “efficiency” of your storage. A very cost-effective way to increase the IOPS/spindle you get out of your existing storage is to install Virsto. Relative to native hypervisor thin provisioned virtual disk technology, we’ve seen IOPS/spindle increases of from 3x (for FC disk) to 10x (for high capacity SATA disk). Virsto’s pure software solution installs at the hypervisor level (not in the guest VMs), works with any block-based storage, and generally increases the “efficiency” of your storage by 3x or more.

The business benefits are pretty obvious. You use a lot fewer disks to meet your performance requirements, you’re generating a lot more IOPS/spindle so you can work with thin provisioned disks to save on storage capacity, and you end up being able to support a lot more VMs with any given storage configuration. Using a lot smaller storage configuration to meet your performance requirements also reduces secondary storage costs for things like backup and disaster recovery, saving even more money.

Maximizing utilization of infrastructure you’ve already paid for is a good thing. You’d do it for manufacturing, you’ve done it for servers, why not do it for storage too?

Want your virtual environment to perform? Spend more money on storage.

by Mark Davis, CEO

Posted Saturday, April 23, 2011 in Alternative Solutions 0 comments

Tags: false assumptions, Hyper-V, storage sprawl

Even after a couple years at Virsto, I am still amused at advice to solve performance issues by wasting disk space.

Michael Otey, from SQL Server Magazine and Windows IT Pro, wrote this week a good article with 7 tips on how to optimize SQL Server on Microsoft Hyper-V.

One of his tips is a very common one: to get better storage IO performance out of Hyper-V, use fixed VHDs rather than dynamic VHDs. This is Microsoft's own recommendation for best performance.

But there's a reason why people want to use dynamic VHDs: they save disk space. A dynamic VHD doesn't pre-allocate precious, expensive disk space. It allocates only the space actually occupied by data. Fixed VHDs, on the other hand, will take up 127GB for that VM image, even if the actual content of the image is 12GB.

What we find odd is the suggestion to solve one problem by creating another. Yes, you can get better IO performance on fixed VHDs. But the price you pay is wasting a lot of disk space. How is that good?

With Virsto software, you can get better performance than fixed VHDs while taking up no more space than dynamic VHDs. Not to mention a whole lot of other benefits. Why would you do anything else?

Wow, what a quarter!

by Mark Davis, CEO

Posted Wednesday, April 13, 2011 in News 0 comments

Tags: customers

Not long ago, I blogged about our first year of shipping product.

It was a great first year of delivering production-grade solutions to corporate and clould customers. Such a pleasure it was to have live customers using Virsto to improve VM application performance while they save piles of money on capital and operating expenses.

We've since finished the first quarter of 2011, and all I can say — once again — is "Wow!"

In Q1 of 2011, Virsto:

Signed more customers than in all of 2010
Signed contracts worth more than in all of 2010

Momentum is a powerful force.

CSI: VDI

by Eric Burgener, VP Product Management

Posted Tuesday, April 12, 2011 in Technology 0 comments

Tags: Hyper-V, Virsto VDI, virtual desktops

VDI and CSI

So I’m watching the news the other night and this ad for some new cop show comes on (it was CSI: Abercrombie and Fitch, I think). It looked pretty formula – attractive, fashion conscious 22 year old forensic pathologist meets dead guy, becomes fascinated with dead guy’s life, and solves the mystery of his untimely demise, all in an hour. Formulas can be convenient, since they tend to be based on past, proven experiences, and Hollywood is a perhaps annoying example of that. I tend to day (or night) dream about storage a lot when I’m winding down at the end of the day, and got to thinking about how formulas apply in an area I’ve been spending a lot of time on lately: virtual desktops.

If you’re about to embark on a virtual desktop infrastructure (VDI) project, what formula are you using to determine the storage configuration you need on the back end to meet your performance requirements? Do you have a variable in there that accounts for the storage performance degradation you’re going to experience when you use a standard, SAN array as the storage back end for a VDI configuration? Does that variable assume that you’re going to get at least 30% fewer IOPS out of your storage configuration than you have been getting with that same array when it’s attached to physical servers? Have you budgeted for the additional storage hardware (spindles, solid state disk (SSD), etc.) that you’ll need to bring that performance back up to where you need it to be?

If not, you might want to consider the following:

Anytime you have virtual machines (VMs) residing on a physical server, the I/O pattern that that server generates is going to be much more random and write-intensive than it ever was when it was just a physical server running one application
Current disk technologies, including both spinning disk and SSD, perform at their worst with very random, very write-intensive workloads – this is the phenomena that results in the storage performance slowdown in virtual environments because rotational latencies and seek times start to dominate data transfer times
The more VMs that reside on a host, the more random and more write-intensive the I/O pattern becomes, leading to the realization among graduates of the “school of hard knocks” that you need a bigger (translate: more expensive) storage configuration in virtual server environments than you do in physical server environments, and you need an even bigger (translate: even more expensive) storage configuration in most VDI environments

Bottom line: make sure you test storage performance with a small, VDI pilot project so you can start to get an idea of what the “storage degradation variable” and storage budget will need to be for your environment.

You can address this problem in one of two ways: throw hardware at it or throw software at it. Throwing hardware means buying more spindles and/or buying faster storage technology (FC instead of SATA, SSD instead of spinning disk, or enterprise-class storage arrays that use components like PAM modules, etc. to help speed up storage performance). Prepare for some heavy lifting in the storage budget area if this is the approach you plan to take, but at least now you won’t be surprised.

The other way is throwing software at it. If an approach that increases the IOPS you get out of your existing spindles by 300% - 500% (regardless of whether they are spinning disk or SSD), works with any heterogeneous block-based storage, and is a simple install at the hypervisor layer (no software in individual VMs or virtual desktops) sounds interesting, you might want to check us out at www.virsto.com. We’ve got other interesting things to say about thin provisioning and snapshot/clone technology that again, are driven by software and available on top of any heterogeneous storage, all of which are available with a single virtual disk type: the Virsto vDisk. Count on bringing your cost/desktop back into the range that will get your VDI project moving again.

Who knows? Instead of the next big show being about autopsies of derailed VDI projects (CSI:VDI?) it might be more along the lines of “Are You Smarter Than A Physical Server Storage Analyst?”…!

Virsto for Virtual Desktops (VDI)

by Alex Miroshnichenko, CTO

Posted Tuesday, April 05, 2011 in News, Technology 1 comment

Tags: Hyper-V, Virsto VDI, virtual desktops, virtualization, VM I/O blender

At the risk of blowing my credibility in the first sentence of this post I am making the following statement: I believe 2011 will (finally) be the year of VDI. After so many false starts and false promises, what has changed to make this year the year VDI gets real momentum?
I believe the key factors are now in place that will drive the market to seriously consider, and successfully implement VDI this year:

Windows 7 Refresh – The positive market reception to Windows 7, and the end of support for Windows XP, provides a compelling event for IT organizations to consider migration options.
VDI solutions have continued to mature, and now support key end user requirements such as high performance network graphics, PC-over-IP delivery, and hybrid desktop delivery options across desktop and mobile devices, with acceptable application latency and performance.
The cost per virtual desktop is finally getting driven down to levels that compel IT organizations to try this new VDI delivery model, as the cost savings with virtual desktops are finally real and large enough to challenge the status quo.

Delivering a lower cost per virtual desktop is the final, critical piece to making this the long-awaited, much anticipated “year of VDI”.
VDI discussions tend to focus on the sexy topics like high performance network graphics and catchy terms like PC-over-IP. If we look beyond these important subjects and focus on the drivers of cost for virtual desktops, the majority of these costs come from the infrastructure requirements needed to stand up a VDI environment.

A VDI infrastructure is essentially a pool of virtualized servers running VERY large number nearly identical virtual machines. When I say a VERY large number I mean that it is not unusual to have more than a hundred virtual machines per host as opposed to a 10-15 VMs for a common server consolidation scenario. With Windows Server 2008 R2 SP2 Hyper-V now supporting dynamic memory, the VM density required for cost-effective VDI is ready to go.

However, the lion’s share of the cost of the supporting infrastructure is still dominated by the cost of storage. And the cost of storage has little to do with the amount of storage actually required, and everything to do with provisioning to deliver storage performance for the VMs running VDI.

This unpleasant storage performance requirement comes as a surprise to a lot of people who have tried to deploy large-scale VDI implementations.

Hundred virtual machines running on a single host create a severe case of what we at Virsto call the “I/O Blender”; many VM’s acting as if they own all of the host’s resources, not playing nicely with all the other VM’s, grinding I/O throughput to a halt and dramatically reducing overall performance as a result. What makes this problem especially acute is the fact that I/O performance is dominated by small block write operations in the VDI use case. The result is the worst nightmare for modern storage systems and a huge storage price tag for users.

Storage hardware vendors are ready to offer their standard solution: buy more expensive storage to over-provision the environment and try to deliver enough performance to solve the I/O bottleneck. Since server virtualization has taken off, the standard way to solve this performance problem has been to either buy systems with more disk spindles; or pay for the solid state storage (SSD) – which despite the recent reduction in price is still orders of magnitude more expensive than traditional spinning disk solutions. With either approach storage costs will blow your VDI project budget: it is not uncommon for storage to account for $300-$400 per virtual desktop in VDI implementations.
The magic number for cost per virtual desktop seems to be around $500 for successful implementations. With current storage approaches VDI is a non-starter for most IT organizations.
It does not have to be that way. Your existing storage – or inexpensive commodity disk storage – is perfectly capable of handling the challenges of VDI workload provided it is controlled by storage virtualization software designed to handle storage in virtualized environments - Virsto Software.

Over the past 6 months we have been working on software optimized to solve the specific storage challenges in the VDI use case. I am proud to announce that we have succeeded. Virsto VDI changes the economics of storage in virtual desktop deployments running on Hyper-V by optimizing the I/O in virtualized environments, supporting highly consolidated environments, while reducing the storage requirements by up to 90%. Virsto supports Enterprise-scale VDI deployments by delivering tens of thousands of high performance snapshots, Clustering and High Availability fail over through Microsoft DPM, and improves VDI provisioning with support for instant bulk provisioning and grouping operations through Microsoft System Center VMM. By introducing Virsto into your VDI POC and getting dramatically more out of the storage you already have, the storage component of your cost per virtual desktop is dramatically reduced.
I think this really could be the year of VDI.

Outperforming VMware: Virsto and Hyper-V

by Eric Burgener, VP Product Management

Posted Monday, March 07, 2011 in Technology 0 comments

Tags: Hyper-V, performance, Virsto One, virtualization, VMware

I recently blogged about how Virsto and Hyper-V combined give you a virtualization platform that is data center ready and compares very favorably on key metrics like performance, storage capacity utilization, and snapshotting to VMware. Let’s take a little closer look at the performance comparison.

We’ve run tests internally comparing Virsto vDisks (which is our version of a VHD) against native Microsoft fixed disks, against native Microsoft dynamic disks, and against VMware thick and thin VMDKs. We offer slightly more performance (it varies between 15% and 30%) than fixed disks and thick VMDKs but give you all the advantages of thin provisioning, high performance snapshots and instant provisioning. We offer 3x – 4x the performance of dynamic disks, and roughly the same against VMware thin VMDKs, again with all the advantages of thin provisioning, high performance snapshots and instant provisioning. And the performance differential gets even greater with Virsto if you plan to use snapshots. In both the native Hyper-V and VMware environments, performance degrades as more snapshots are created and retained. Not true with Virsto, so the performance advantage improves even further. Clearly, Hyper-V with Virsto can significantly outperform native VMware, assuming the same hardware configuration, in situations where you want high performance, thin provisioning, and snapshots at the same time.

One other comment I’ll make. Both Hyper-V and VMware support shared ownership of virtual hard disks at the VM rather than at the LUN level. Hyper-V uses cluster shared volumes (CSVs), while this feature is native to VMFS. This feature is important to provide for granular failover of VMs, something that has value not only for HA but also for workload balancing. Virsto vDisks support this same capability – shared ownership at the VM level. We’ve seen some performance impacts with CSV use that I’ll blog on in a later post, but I want to underline that vDisks support this capability without any of the performance issues we’ve seen with CSVs.

If you want high performance, both against your data and for your snapshots and clones, and thin provisioning together, there is no doubt that Hyper-V with Virsto installed gives it. And that combination handily outperforms VMware.

Wow, what a year!

by Mark Davis, CEO

Posted Tuesday, March 01, 2011 in Categories, News 1 comment

Tags: customers, startups, Virsto One

Just over twelve months ago, Virsto Software released our first product to the world.

February 2010 was the culmination of several years of work to develop this product, yet it also was the commencement of addressing some key challenges in the virtual and cloud computing market. We’re confident it will lead to even more years of success and further evolution.

The impact of Virsto and our technology reaches beyond simply helping companies consolidate their storage and save a few bucks here and there. It is instead a truly essential shift in how companies virtualize their data centers and consume storage.

We are at the confluence of several fundamental trends, including widespread data center virtualization and the growing adoption of cloud computing. For cloud computing to realize its full potential, it is necessary to virtualize the server and the storage elements. Virsto is playing a huge role in making proper storage virtualization a reality and helping the cloud live up to that vision. Already we have customers that are delivering cloud services enabled by the Virsto platform.

Over the span of the past year, we have hit quite a few important milestones:

Signed 20 customer deals worldwide.
Helped these customers achieve efficiencies, performance, and improved operations in ways they previously thought impossible.
Saved our customers $10 for every dollar invested in Virsto software.
Achieved 100% quarter over quarter sales growth every quarter since launch.
Accelerated customer adoption by signing cloud service provider and enterprise customers.
Delivered four production-grade product releases and added dozens of features based on our customers’ input.
Expanded our partnership and product integration, moving beyond our Microsoft BizSpark One partnership to also become a Microsoft Virtualization partner, a Microsoft System Center Alliance partner and a Citrix Ready Partner. We’ve also built relationships with other virtualization, storage, server, and cloud vendors, as well as a number of channel partners we’ll be announcing soon.

We are grateful to our excellent customers like Millard Lumber, SSI, West Gastroenterology Medical Group (West GI) and Sienna who recognized the value that Virsto One could deliver. West GI, for example, had been experiencing server consolidation challenges. When Femi Adegoke, IT director of West GI, came across Virsto, he realized his company had been spending far too much on storage. Since Femi began working with Virsto, he has reduced his storage capacity requirements by 75%. With Virsto, he will be able to get more useful capacity and a 3x improvement in performance from his existing hardware.

Beyond Buggy Whips

I have talked to hundreds of virtualization users. In every single conversation, they invariably mention the new problems that virtualization creates for storage administrators and IT teams.

Despite the cries for help, there still hasn’t been much fundamental innovation taking place. It is mostly a case of treating the symptoms rather than the illness itself. The core requirements of virtualization—the workflow, the way the virtual workload interacts with the storage system, etc.—are still incredibly suboptimal. We’re using buggy whips to try to make our automobiles run faster.

The world of virtualization is still growing at breakneck speed, and we run across people every single day who need help figuring out how to make the virtual environment work and deliver the benefits they need—flexibility, agility in their infrastructure regardless of size, cloud support—and storage systems today are holding people back.

That’s why Virsto is proud to put something into the market that’s ready and is ideally suited to address those fundamental issues in an innovative yet real way.

What’s Upcoming?

We’re proud of what’s been achieved since Serge, Alex and I started in Serge’s garage and my dining room, brainstorming a project worthy of many years’ hard work. But there is so much more to come! In the next year, here are a few things you can expect from Virsto:

Evolution of the use cases for Virsto: We have already experienced traction among hosting companies and enterprises. We haven’t talked about it much publicly, but there are also customers employing Virsto software in a virtual desktop environment. You’ll be hearing more about Virsto in VDI very soon.
Emergence of new platforms: People tend to think of Virsto as a Microsoft Hyper-V company. But you want to know a secret? Our software ran on another platform in our lab before we ever made it run on Hyper-V. Still does. In fact, it runs on a couple other hypervisors in our lab right now. You’ll see us deliver our solution on at least one other virtualization platform this year. Just as we always promised.
Expansion of our partnership program: We’ve been quietly investing in relationships with a number of strategic and channel partners. You’ll hear a lot more about these in 2011.

People are yearning for a whole new approach, a different way of thinking about what storage means in a virtual world. That’s why we’re here. We’re proud of the problems we’ve solved so far and are eager to tackle others that are still out there. Buckle up for another exciting year.

Virsto and Hyper-V: A Compelling Alternative in the Data Center

by Eric Burgener, VP Product Management

Posted Friday, February 25, 2011 in Technology 0 comments

Tags: Hyper-V, Microsoft, VMware

It was clear when Hyper-V R2 first shipped in 2009, that Microsoft was directly coming after VMware. Since then, Hyper-V R2 has evolved at a rapid rate with major new feature additions, the latest key one being the release of dynamic memory in Hyper-V R2 SP1. There are still some challenges when deploying Hyper-V in the type of complex environments required in larger production environments though. If you’re looking to enhance your experience with storage performance, storage capacity consumption, provisioning times, snapshot backups with Windows Volume Shadowcopy Services (VSS) and Microsoft System Center Data Protection Manager (DPM) or cluster shared volumes (CSVs), then you will be interested to know more about Virsto.

Virsto provides a software plug in to Hyper-V R2 that integrates seamlessly with System Center management tools through an MMC snap-in and puts Hyper-V into “hyper drive” in 4 areas:

Achieve dramatic performance gains: Virsto increases the number of IOPS you can push through your existing disk spindles by as much as 3x – 4x.
Make storage capacity consumption much more efficient: Virsto decreases storage by using thin provisioned disks.
Accelerate VM provisioning: Virsto delivers provisioning via snapshot/clone technology that works with heterogeneous storage, supports an unlimited number of snapshots without consuming any additional storage capacity and introduces no performance degradation.
Enable high Performance HA: Virsto enables most of the “HA” functionality of CSVs to allow shared access at the volume level without any performance, snapshot backup, or management limitations.

Virsto achieves these benefits by implementing an innovative virtual storage layer that runs on any heterogeneous, block-based storage and is completely transparent to administrators and applications after initial installation. What these features provide in terms of bottom line benefits to our customers running in production today include:

Improve the ROI of virtualization projects. Lowering the overall cost/virtual server by significantly increasing the VM density they can support with their existing storage infrastructure.
Jumpstart stalled VDI projects. Enabling VDI deployments where, prior to using Virsto, high storage costs had made the cost/desktop prohibitive and effectively cancelled the project.
Defer incremental storage purchases. Delaying the need to purchase additional storage hardware (in cases where these customers already owned the storage) by making much more efficient use of existing storage resources.
Reduce administrative overhead. Saving hours and in some cases days in provisioning times by enabling instant snapshotting that consumes no additional storage capacity.
Enable flexibility with lower cost storage. Allowing customers to build highly available virtual environments that would otherwise have had to leverage CSVs without requiring the purchase of enterprise-class storage to address performance and snapshot backup limitations.

Analysts are starting to comment on our technology as well. Storage Switzerland took a look at our technology recently and their takeaway was clear: “Virsto also makes Hyper-V more relevant and capable. They may by themselves cause a data center to more strongly consider Hyper-V than they did in the past.” You can check out the rest of the blog here. You can also see what some of our production customers are saying in case studies on our website (hosting provider SSI, etc.).

If you want a data center-ready server virtualization alternative to vSphere based on Hyper-V R2, you might want to look into Virsto. Feel free to try Virsto out by downloading an eval here.

Dirty Little VDI Secrets

by Eric Burgener, VP Product Management

Posted Thursday, February 24, 2011 in 0 comments

Tags:

If you’re just starting to think about a VDI deployment, there are a few dirty little secrets that I hope I can help you get in front of. For an issue to qualify as a dirty little secret, I’d say it’s something that isn’t readily predictable up front, and has some real negative consequences down the line. After having gone through the process of VDI deployments vicariously through some of our customers, here’s what I’ve seen on the storage side that seems to meet that definition.

Each desktop that you’re virtualizing may have less than 25GB of data on it, but when you’re virtualizing 500 or 1000 of them, that can be a lot of data that you are now centralizing in one place. On physical desktops, that storage capacity was probably sitting in some pretty inexpensive IDE drives. Most of these desktop systems come with a pretty good size drive (for individual user purposes) and I note that on frys.com I can buy 1TB of this class of storage capacity for a little more than $100. But when you centralize all that storage in the data center, because of a lot of other considerations (performance, high availability, recoverability, etc.) you’re probably putting it on a much higher cost class of networked storage. So I could go from paying ten cents a gig to paying more like $4 - $8 a gig for raw capacity. Maybe this is a surprise, and maybe it’s not, but this next one probably is.

So let’s say you’ve gotten past this first issue, and have gone forward with your VDI project. You’ve bought server virtualization software, chosen hosts and storage, and obtained the necessary VDI software. Presumably you did some planning to determine what sort of storage configuration you would need, using the same general approach you’ve used on physical servers in the past. Profiling the workload (IOPS, MBPS, latency), understanding avg and peak requirements, then looking at availability, performance, and capacity requirements, and coming up with what you need from there.

Well, here’s the surprise. Because of the way type 1 hypervisors like ESX, Hyper-V, and XenServer work, your storage is going to run a lot slower in the average virtual environment than it does in the average physical environment, and here’s why. Multiple VMs running on a single physical host create an I/O pattern that is very write intensive and extremely random, and this is exactly the pattern that will get you the worst performance out of your disks, regardless of what class of device they are. Seek times and rotational latencies really start to dominate, and what we’ve seen in most customer environments is that storage performance can easily take a 50% hit. And if you’re trying to use thin provisioned disks to save on capacity, you’re going to take even more of a performance hit unless you buy a very high end class of storage array that addresses that problem for you.

If you can afford that kind of high end storage, then you may avoid the performance hit due to thin provisioning, but you’re still going to run into the first problem, and end up having to throw more spindles at your storage config to get the IOPS you need. SSD could be another option, but if your workload is very write intensive, that’s probably not going to help much.

Where I’m headed with all this is that the legacy storage architectures that worked pretty well with the one server/one application model that has been in vogue over much of the last 20 years probably isn’t going to cut it in the virtual world. And virtual is where we’re headed.

There’s another option to consider here. At Virsto Software, we’ve got a software plug in for hypervisors that sets up a storage virtualization layer, compatible with any heterogeneous block-based storage, that turns extremely random I/O patterns into sequential I/O patterns as they hit physical disk, which is how you’ll get the highest performance out of your disks, whatever class they are. This layer is transparent to applications, integrates seamlessly with the hypervisor and its associated management tools, and does two things right off the bat: it increases the number of IOPS you can push through your existing spindles by 3x – 4x, and it reduces your storage capacity consumption (through software-based thin provisioning). There’s no hardware to buy, and you’ll end up needing a lot less than you thought you did.

This is a relatively new technology, but it is running in production in a number of accounts, and it’s just one more option you have in dealing with unexpected surprises for VDI implementations. By the way, this same storage performance issue exists in server (not desktop) virtualization environments as well, and this plug in has the same benefit.

As VDI deployments get bigger, there’ll probably be a few more surprises we as an industry run into. But I hope knowing about two common ones helps you in your planning a bit.

Cost of Storage Infrastructure Will Have A Huge Impact on the Evolution of Cloud

by Eric Burgener, VP Product Management

Posted Monday, February 21, 2011 in Technology 0 comments

Tags: cloud, virtualization

Issues like performance, ease of provisioning, security, and end user pricing will clearly be key issues, but an issue which underlies all of those is the cost of building out public and private cloud infrastructure.

Because of the flexibility cloud environments demand, virtualization technology is really the only way to deliver on the Cloud approach. Cloud providers will have to focus on creating the most cost-effective virtualized infrastructure to meet their particular customer requirements so that they can price competitively while still generating acceptable margins.

What I think will happen here is that a “standard operating environment” (SOE) will evolve over time that will be quite similar across different cloud providers. This SOE will be built around particular technologies, not particular vendors, that make more efficient use of existing hardware.

Virtualization of servers is already a prerequisite. The single component in that stack that has the biggest impact on infrastructure cost is storage, and traditional storage architectures are particularly inefficient in server virtualization environments. The next wave of virtualization is now focused on storage virtualization, and specifically storage consolidation. Just as the primary driver of server virtualization was server consolidation, the new imperative for IT is to drive down storage costs using new approaches to consolidate storage.

One promising approach to drive down storage costs relies on a new storage virtualization layer that is designed from the ground up to deal with the extremely random I/O patterns that virtual environments generate.

A layer like this, implemented in software in the hypervisor, is easy to deploy and preserves server, storage, and networking hardware as well as hypervisor platform choice, resulting in a cost/server or cost/desktop that traditional storage technologies, which are optimized for the one server/one application model that has been so prevalent over the last 20 years, just can’t touch.

The storage consolidation savings will come from generating more IOPS/spindle of existing storage capacity, better utilization of solid state storage capabilities through tiering, and consuming less storage capacity through thin provisioning.

These factors will increase virtual machine density in these environments and reduce price per unit of computation and storage.

Hosting more servers or more desktops per virtualization host and/or per TB of storage capacity are key variables to manage in keeping virtualization infrastructure costs down. The SOE that evolves for cloud environments will move away from traditional storage architectures as it chases that goal.

Dynamic Memory will highlight storage performance problems

by Eric Burgener, VP Product Management

Posted Friday, February 11, 2011 in Technology 0 comments

Tags: Hyper-V, performance, virtualization, VM I/O blender

Service packs are coming out Feb 22 for Windows Server 2008 R2, and one of the key new features is dynamic memory (see Jon Brodkin's article). By boosting how efficiently Hyper-V makes use of memory, this should boost VM density on a given host for a given set of resources. This claim tracks pretty closely with one we've been making for storage resources - that with a simple hypervisor plug in we'll let you support more VMs on a given storage configuration by significantly increasing the performance you get on a per spindle basis.

Memory capacity is probably the main limiter of VM density, but storage is right behind it in terms of importance. How many of you haven't noticed that you need to add spindles to your existing storage configurations once you start hosting workloads in virtual environments? Anyone...? This issue doesn't really have anything to do with main memory, but has everything to do with how hypervisors in general (not just Hyper-V) handle I/O. You might have heard the term "VM I/O blender", and its an apt description of what happens to I/O in a host running multiple VMs. The extremely random I/O pattern pretty much means you get the worst performance your underlying disk spindles can support because rotational latencies and seek times start to dominate disk access times. With such a random I/O workload the hypervisor doesn't really have the opportunity to performance optimize how data is written to disk. What's needed is a storage architecture that is built to handle the I/O patterns common in virtualized environments, not one designed for the good old days of "one application/one server" that have dominated the client/server computing model. This problem exists in virtual server and virtual desktop environments, and the problem tends to be worse the more write-intensive the environment is.

How does this relate back to Microsoft's release of Windows Server 2008 R2 SP1? First, if you're having performance and/or density problems on Hyper-V, memory might be your problem, and it might not. If it isn't, then storage performance probably is, and you might want to check Virsto out. If main memory was your problem, and dynamic memory helps, that will probably pretty much just move the challenge downstream to storage.... and you might want to check Virsto out.

So I guess the bottom line here is that if you want the best performance you can get out of your existing storage configuration, you'll need to optimize along the entire I/O path, not just the main memory. The gains you'll get on Hyper-V from implementing both dynamic memory and Virsto will be pretty impressive, so don't just do one or the other... do both. You can see from case studies on our site that a lot of our customers get 3x or more the storage performance they were getting out of Hyper-V before they installed us while actually using fewer disk spindles. If you want to see what this might mean in your environment, you can download a trial version of our software for free at www.virsto.com.

You want Hyper-V to outperform ESX?

by Alex Miroshnichenko, CTO

Posted Tuesday, February 08, 2011 in Technology 0 comments

Tags: bloggers & media, Hyper-V, performance, VMware

Deepstorage.net, a third party dedicated to providing objective analysis, testing, and reviews posted a N etwork Computing blog recently that discussed the performance difference (in terms of IOPS) they had seen when testing Hyper-V and VMware ESX against the same backend storage configuration (a SnapSAN 2000 from Overland Storage). Overland had hired Deepstorage.net to run the tests, but that wasn’t what the blog was about. In his blog, Howard Marks commented on the fact that in these identical tests ESX seemed to consistently outperform Hyper-V, leading him to conclude that ESX had a cleaner I/O path than Hyper-V. Deepstorage.net’s multiple workload benchmark showed Hyper-V pushing 733 aggregate IOPS vs ESX’s 955 IOPS, indicating roughly 25% better performance with ESX.

At first I was surprised that the original post and the testing report available on the Overland site did not indicate one of the most important consideration: what kind of storage configuration was used. In an apparent response to my comment Howard Marks clarifed that he was using thinly provisioned VMDKs on the ESX side and dynamic VHDs on the Hyper-V side - so the testing was done in a somewhat comparable scenarios. However I was surprised to find out that in both case the thinly provisioned virtual disks were fully expanded prior to the actual benchmark run. In my opinion it defeats the main purpose of configuring the thinly provisioned objects and confuses unintitated readers.

But that’s not even the real point of this. The point is, native storage options in hypervisors today are pretty pathetic, regardless of whether that’s Hyper-V or ESX. It’s because the storage layer in use - be that VMFS or NTFS - was never designed for the virtual world. The very random, write-intensive I/O patterns so common in virtual server environments pretty much kill physical disk performance. And the hardware vendors are the ones benefiting from that – it means they can sell you a lot more spindles. If you like that, great. If you don’t, you might want to know about what we’re doing.

We’ve got a true virtualization storage software solutuion that is designed specifically for the types of I/O patterns you see in virtual environments. You want Hyper-V to outperform ESX? Based on what we’re seeing at customer sites and in our own internal testing, we’ll have Hyper-V routinely outperforming ESX while we use a lot fewer disk spindles and less storage capacity. Our storage objects (we call them virtual VHDs) look like VHDs to Hyper-V, but they outperform fixed disks by 3-4x and use on average 30% fewer disk spindles. This all assumes the same storage back end for the comparisons (Virsto vs native Hyper-V) in tests we’ve run. This definitely beats the puny 25% perfromance difference reported by Deepstorage.

Sound interesting? Download our software today and check it out. And welcome to the world of 21st century storage…!

Cloud, storage, and venture capital opportunities

by Mark Davis, CEO

Posted Monday, January 10, 2011 in Software Business 2 comments

Tags: cloud

The folks at VC Journal asked me to share some thoughts on investment opportunities in the cloud. Given the recent large VC returns in the "cloud storage" space (I use the term loosely), I thought it might make sense to look at the topic from that point of view. Unless you're a VC with a subscription to VC Journal, you won't be able to read it on their site, so I'm pasting what I submitted below. Curious to hear your feedback.

The future of cloud storage venture investments

Storage sure looks sexy to venture investors these days. The acquisitions of 3PAR (HP), Isilon (EMC) and Compellent (Dell) add up to more than $5.5 billion. All three were storage hardware companies of the same vintage, about a decade old, and big wins for their VC investors. Looks like virtualization and cloud storage are the places to invest, no? Well, I sure think so, and placed a huge personal bet three years ago by starting a company precisely aimed at these markets.

It’s interesting to note that all three deals were partially justified by acquirors, financial analysts and pundits as plays in the new virtualized, cloud data center. Yet the core technologies touted to justify lofty valuations were developed before the emergence of virtualization let alone cloud computing.

When they were founded, none of these three storage companies were designing for a world of virtualized clouds. None of them redesigned their products from the ground up for these new markets. They did change their marketing messages, to be sure. With this repositioning, they achieved cachet, and I cheer for their successfully grabbing the brass ring.

As a venture investor, the question is what storage plays will pay off 5-10 years hence, when virtualization and cloud become mainstream?

3PAR + Isilon + Compellent = Game Over?

Will the new owners of 3PAR, Isilon, and Compellent take all of the market, in which case new startups needn’t apply? History and technology trends suggest not. There are plenty of technical obstacles yet to be solved in this space. Proprietary hardware systems architected ten years ago, now captured by giants who inherently aren’t as innovative as startups, aren’t likely to be the ultimate answers.

One thing is certain: cloud computing and virtualization are fundamentally different. It is clear that new approaches to storage are going to be game changers in enterprise and cloud virtual computing. It’s a very good bet that the big winners a few years from now will emerge from a whole new generation of startups.

VCs who drive by rearview mirror bewilder me. Why make investments that won’t become liquid for years based on today’s hot exits? Will the next big plays in storage for the virtual cloud be 3PAR part deux – more monolithic, proprietary storage hardware?

There will surely be more significant storage exits as the cloud and virtualization markets develop. What is not so obvious is that they will be proprietary hardware plays.

In the computer business, value has migrated from hardware to software. Storage has lagged in this regard by a decade or so, but the transition is tangible. That’s why my bet in the next generation storage market is on software.

The Case of the Pending VM Snapshot Merge

by Serge Pashenkov

Posted Friday, December 10, 2010 in Alternative Solutions 0 comments

Tags: Hyper-V

Over on the Microsoft TechNet blog, Jeremy Hagan speaks of the woes of native snapshots in Microsoft Hyper-V:

"When a snapshot is deleted, the virtual machine must be powered off, not merely rebooted, in order for the data in the differencing disk to merge into the parent VHD. The VM must remain powered off until the merge is complete. If the machine is booted back up again before the merge is completed then the merge process stops until the machine is powered off again. The bigger the AVHD, the longer the merge takes. And while SCVMM and Hyper-V manager have an indication in the GUI that a snapshot exists, when a snapshot is deleted it is deleted from the GUI straight away even though the snapshot hangs around until the merge is complete. So if there is a VM in this state there is no visible indication of it."

Whoa. Seriously? This is the way snapshots are supposed to work? I don't think so.

Jeremy then explains how this resulted in problems for his environmnent, and he kindly shared a bunch of code that will alert him to the problems when they happen next time.

My reaction to reading this is, why all the hacking to deal with what seems like a design bug in native VHD snapshots?

Snapshots are supposed to make life easier

Virtualization makes it possible to do many things differently, much more efficient, better, and easier. Enter virtual machine snapshots and cloning. Say you need to repeatedly run a set of tests, and you want the test process to be reproducible. What better way than to create a snapshot of a VM, clone it every time you start a test, and run the test against the clone. Every run starts from the same exact state. Perfect!

Now imagine you did find a problem. VM snapshotting helps again. Take a snapshot of an erroneous state so that it can be reviewed later. Even better!

Things go pretty well for a while. Then you either run out of space, or simply become overwhelmed with the number of snapshots to keep track of. So you want to cleanup, which means you need to delete some snapshots. As Jeremy's TechNet post has shown, that can be quite a painful situation.

It makes you feel like all the air went out of the wonderful balloon you were using. Those snapshots are useful until you have to delete them, then the trouble begins.

An alternative is to use Virsto One snapshots. Virsto snapshots and clones are designed to not only perform well when they run, but when they are being deleted. With Virsto, there are no limitation like Jeremy and other Hyper-V users have to deal with. With Virsto, a VM could be happily running while its snapshots are being deleted. No need to power down the VM.

Another case to think about: backup. Isn't it natural to use snapshots for backup? I mean you can take a snapshot of a VM and use it for backup while it runs. So far so good. But wait a minute. After the backup is done you probably want the snapshot deleted, right? Is it ok for the VM to be powered off and remain powered off until the snapshot delete is completed?

I must say it didn't occur to us when we were designing Virsto One that the kind of limitations Jeremy wrote about would be acceptable! There are a number of use cases for snapshots in a VM environment, and all of them tax traditional technologies. That's why we designed Virsto One to be a lean mean snapshotting machine!

More developers wanted!

by Serge Pashenkov

Posted Friday, December 03, 2010 in Technology 0 comments

Tags: Hyper-V, people, startups

We are extremely proud of the development team at Virsto Software. It is amazing how powerful a platform we have delivered in such a short time, with such a tight little team. It has been a pleasure to be a part of it.

We've built a quite impressive core product on Microsoft Hyper-V, and the reaction from customers and partners is really exciting.

As good as the team is, we need more people. We have a number of interesting projects in our roadmap, and we need help!

This is where you come in. If you are a world class storage I/O software engineer, or know someone who is, please get in touch with us. We're looking for several developers who fit the following description.

Core I/O engineer

Requirements:

Multi-platform software development experience, specifically with Windows and Linux.
System level experience with modern operating systems, specifically in high-performance I/O subsystem, file system, disk I/O.
Proficiency with C and C++ languages, and multi-threaded programming.
General understanding of modern I/O architectures, interconnects, and protocols - one or more of SCSI/iSCSI/SAS, SATA, FibreChannel, NFS, CIFS.

Ideal candidates would:

Be familiar with server virtualization, specifically Hyper-V, Xen, and Vmware.
Be experienced with networking and distributed systems programming.

Definite plus would be:

Familiarity with WMI, MMC, C#
Some experience with Linux FUSE

We pride ourselves on a highly communicative staff that deeply cares for our employees. Competitive packages, great benefits, a fully stocked kitchen and an open door policy to hear input from anyone at any level make Virsto a fantastic place to work.

Using Hyper-V differencing disks with VDI

by Alex Miroshnichenko, CTO

Posted Monday, November 01, 2010 in Technology 0 comments

Tags: excessive storage spending, Hyper-V, performance, Virsto One, virtual desktops

MSDN Last week over on the MSDN blog, an article was posted about the virtues of Microsoft Hyper-V differencing disks in VDI environments.

Recently, I've spent a fair amount of time testing storage configurations for VDI on Hyper-V. Based on that experience, I believe it is worth commenting on the MSDN article. Due to the architecture of Hyper-V differencing disks, there are some serious performance issues that must be considered.

Using Hyper-V differencing disks for VDI image provisioning may solve the problem of rapid creation of several copies of a master image template. However the performance issues with differencing VHDs are much worse than one might conclude from the MSDN article.

Differencing VHD (DVHD) performance is noticeably worse than that of a fixed VHD. The VHD performance paper referenced in the post measured the performance of fully expanded DVHDs, even though write performance of unexpanded DVHDs is much slower. Making any conclusion on the performance characteristics of either non-expanded dynamic VHDs or differencing VHDs based on that paper is incorrect at best and could be disastrous in practice. In the case of VDI, dynamic VHDs are almost always in the non-expanded state.

VDI IO + dynamic, differencing VHDs = likely performance issue

VDI IO loads are dominated by small random write requests. This is exactly the worst case scenario for differencing VHDs. The article on MSDN mentioned this, but glossed over the point far too lightly for my taste.

VDI IO patterns will lead to massive extensions of differencing VHDs and will result in a significant degradation of overall VDI performance. DVHD extensions are done at the granularity of at least 1MB at a time, while the actual IO causing the extension is typically between 8KB and 128KB. The difference between the IO size and the extension page size has to be filled by writing zeros, which leads to significant increase in IO traffic, even further reducing net performance.

A single host may support on the order of a hundred VDI instances, so any real world VDI deployment will have to involve more than a single Hyper-V host. So differencing VHD files will heve to be deployed on a CSV volume, which unfortunately will lead to additional performance problems. Any differencing VHD extension becomes a volume-wide operation which is routed through a single CSV coordinator. Therefore, I believe VDI performance of dynamic and differencing VHDs with CSV is simply below the minimum acceptable level.

Virsto One = higher performance + lower cost

Using differencing VHDs for VDI image management is possible for simple small scale demonstration purposes, as it does speed up the setup process. However, I don't think differencing VHDs will prove acceptable for any practical production VDI configuration.

Virsto has conducted a performance comparison study of fixed and dynamic VHDs in a VDI environment. We have demonstrated conclusively the choice between fixed and dynamic (differencing) VHDs is the choice between unacceptably high storage capacity (and therefore cost) versus unacceptably low performance. Not a palatable tradeoff.

Have your cake and eat it too! The only practical option for deploying scalable VDI configurations with Hyper-V while keeping storage costs under control is to use Virsto One. Virsto's patent-pending technology delivers outstanding performance at the same time as saving space through thin provisioning (à la differencing VHDs) and snapshotting (à la differencing VHDs).

With Virsto, IO performance is as good or better than fixed VHDs, and at least several times faster than dynamic VHDs. At the same time, the amount of disk space used with Virsto one is even less than with differencing VHDs.

See, you can have your cake and eat it too.

Vote for VM storage discussion at Tech•Ed Berlin 2010

by Mark Davis, CEO

Posted Friday, October 22, 2010 in Events 0 comments

Tags: EMC, virtual desktops

TechEd Berlin 2010 Are VM storage problems on your mind? They're always on ours.

If you're interested in Birds of a Feather sessions at Microsoft's Tech•Ed in Berlin next month, Virsto is offering to sponsor two discussion forums.

But we need your vote! Please click the links below to ensure these sessions happen.

BoF #1 is jointly proposed by EMC and Virsto:

Reducing cost while improving performance when deploying server virtualization

Provisioning new VMs should not take hours. Technologies including thin and rapid provisioning can reduce costs while improve performance. Using these technologies you can rapidly deploy new VMs while reducing the overall deployment time down from hours to minutes. Technolgies exist today that can help reclaim 90% of an enterprise's time and space for a more efficient data center. Vote here!

Bof#2 is offered for your pleasure by Virsto:

The virtual desktop (VDI) storage challenge

VDI implementations are characterized by write intensive random I/O patterns. Vasts amount of expensive high performance storage is typically deployed to properly sustain a sizable VDI deployment. As a result the cost per virtual desktop instance is often prohibitively high - the storage costs being one of the main obstacles to the wider VDI acceptance. The BOF session will cover various software approaches to drastically reduce the size and cost of storage while maintaining and even improving performance of scalable VDI implementations. Vote here!

We're looking forward to seeing you at Tech•Ed in Berlin!

A hypervisor IO performance benchmarking recipe

by Alex Miroshnichenko, CTO

Posted Friday, October 08, 2010 in Alternative Solutions 5 comments

Tags: Hyper-V, Microsoft, performance, Virsto One, virtualization, VM I/O blender, VMware

This post is in response to a comment from Eric Gray to my post about hypervisor IO performance. I'd have put this all in a comment to his comment, but our blogging software doesn't like responses with rich formatting (arrgh) and my answer is a bit hard to read without some formatting.

Eric asked for more details about how to run certain IO benchmarks, so I'll provide that here. Eric also asked Virsto so submit our benchmarks to VMware for approval; we'll save our response to that for another post.

As you undoubtedly already know Eric, disk IO workloads in virtualized environments are typically random, have a higher proportion of smaller request sizes, and are more write intensive. So the benchmarking of a hypervisor IO subsystem is relatively straightforward for those with the right expertise. I thought I gave enough hints in my earlier post, but here is a more detailed recipe:

Take your favorite disk IO benchmark utility, say Iometer, and configure it to run a server style profile load (something like 8k to 64k block size, 50/50 read/write, random IO pattern).
Configure your ESX host to run multiple VM instances, each containing an identical instance of the configured benchmark.
Now run 1, 2, 4, 8 and so on VMs simultaneously with active benchmarks in each VM.
Observe how the COMBINED throughput of ALL benchmark instances degrades as you increase the number of VMs running. This is called the VM IO Blender, which effectively limits the density of VMs that can be run per server.
Go back to step 2, and reconfigure your VMs to run on thinly provisioned VMDKs.
- Repeat the runs and observe the performance of the thinly provisioned VMDKs.
  (Hint: the results won't be impressive.)
Go back to step 3 and activate the linked clones on your VMDK.
- Repeat the runs and observe linked clone performance.
  (Hint: the results will be even worse.)
On the same hardware, repeat the steps 1-6 above with Hyper-V as the hypervisor and Virsto as the storage layer; collect the data; draw your own conclusions.
- Note: Virsto virtual disks are always thinly provisioned, even though they perform as if thickly provisioned, so that will save you one step.
- Use Virsto clones when comparing performance to VMDK linked clones.
Go step 1 and replace the benchmark profile with something like 80% to 90% writes, 8k random IO which is more characteristic of VDI workloads. Repeat steps 1-7.
Draw your own conclusions.

Of course this is an outline, and it will take some time, maybe a few days to go through the complete exercise. But I think the results awaiting you at the end justify the effort.

I will be happy to work with you on proving the point in your environment.

Thank you again, and best regards,
Alex.

I/O performance: Not all hypervisors deliver the same results

by Alex Miroshnichenko, CTO

Posted Sunday, October 03, 2010 in Alternative Solutions 2 comments

Tags: bloggers & media, Hyper-V, no dupe, performance, storage sprawl, Virsto One, VM I/O blender, VMware, Xen

I/O performance

Over on the Technet site, VMware’s Eric Gray asked a couple followup questions to my guest post on the Microsoft Virtualization team blog. I can’t give a complete response in a brief comment, so I’m moving the discussion over here to the Virsto blog.

My post extolled the virtues of Dynamic Memory, a feature of the upcoming SP1 release of Microsoft Windows Server 2008 R2. Why would a storage guy love dynamic memory? Because better memory efficiency further exposes virtualization users to another big VM performance bottleneck, namely storage I/O. Since that’s one of the problems Virsto One deals with so well, at Virsto we are fans of efficient memory performance!

Now, since certain vendors (I didn’t name names) like to crow about how much higher performance their platform is than Microsoft Hyper-V, and since Virsto had recently run some storage I/O performance tests, I couldn’t resist pointing how Hyper-V plus Virsto One can be vastly faster than alternatives.

Eric, who I had never met (nice to meet you Eric), responded by asking me to explain the benchmark upon which I based my assertion.

“The first rule of VMware benchmarks is that you don’t talk about VMware benchmarks”

(That's a quote from another response to Eric's comment.)

Now Eric, being the unofficial chief competitive blogger for a top virtualization vendor, knew very well that if I was referring to his employer’s product, I couldn’t answer his question without violating the EULA, which prohibits publishing benchmarks without the vendor's approval.

As much as I think EULA benchmarking restrictions are deplorable, I do believe in software licenses. More importantly, Virsto’s lawyers do, so I was stuck. I couldn’t publish benchmark results without the hypervisor vendor’s approval.

Trust me, this vendor’s marketing department is not going to like the numbers. So I’m not even going to think about getting on bended knee to ask for their approval.

A little side rant

I hope you’ll forgive me, dear reader, for taking a moment to echo the comments of my friend Simon Crosby and others.

Not allowing users (or other vendors) to publish benchmarks using your software is not a sign of strength. Such practices are unfortunate for end users and the industry at large.

I’ve free for all worked for large vendors who are defensive about the performance of their products being compared publicly in uncontrolled (by the vendor, I mean) benchmarks. I understand all the arguments against allowing a free-for-all information exchange out in the light of day. Frankly, I find those defenses unconvincing.

In my experience, most of the people who are employed by vendors that put benchmarking handcuffs in their EULAs are embarrassed by such restrictions, but of course they can never say so. The first rule of EULA benchmark restrictions is you don’t talk about EULA benchmark restrictions.

One “good” reason not to allow benchmarks

Having said that, I will make one concession.

Every sophisticated software vendor knows that letting people without adequate expertise run their own benchmark is asking for trouble. Few people really have the ability to run meaningful benchmarks. It is hard to do, requires a lot of rarified knowledge, takes a lot of time, and can cost a lot of money.

In the storage world, a common example is the IT administrator who benchmarks I/O performance with the dreaded dd command to copy a big file from here to there. This is probably the most ineffective, misleading, and non-real world benchmark for just about any application workload. It is a really bad benchmark for hypervisors. Yet people use it all the time.

It is painful to be a vendor having to defend benchmarks that are run by people who don’t know what they are doing. So I do have empathy.

However, we at Virsto know a thing or two about storage performance benchmarking. Some of us have been doing it since long before VMware existed.

Back to Eric’s query

On TechNet, I said “I would be happy to show any virtualization user how they can reproduce the results themselves”. When I said that, I wasn’t meaning to say that I’d show Eric. But hey, in addition to working for a virtualization vendor, Eric is probably a user too, so I may as well just answer his question.

So how could a user recreate performance benchmarks that support my claim? It’s pretty easy. I’ll give you just two examples.

There are a couple notable weaknesses that many storage solutions have in VM environments.

One is thin provisioning performance. Thin provisioning, in which storage space is not allocated until data is actually written, is particularly useful in VM environments where VM sprawl results in storage sprawl. There are a lot of thin provisioning products out there, and some perform well. Some don’t. A benchmark tool like Iometer can be used to spew random writes to unexpanded thinly provisioned virtual disk files. Why random? Because that’s the kind of writes that a hypervisor has to deal with in the real world.

Second, you might want to check out how your storage solution performs when you turn on space efficient cloning, which shares common data blocks across multiple VM disk images. To avoid VM storage sprawl, space efficient cloning is an essential tool. Several vendors we know of have a cloning feature (the name of one vendor’s rhymes with “distinct groans”), but caution their customers against using them in production environments because of known performance weaknesses. What is the point, I wonder, of features that you’re not supposed to use in production? That same Iometer benchmarking tool can be used to test I/O performance of any vendor’s clones against Virsto clones.

If you want detailed instructions on how to run the benchmarks, please contact us and we’ll be happy to walk you through it. We’re pretty confident that, if the benchmarks are run properly, you’ll be interested in using Virsto One in your VM environment.

We stand by our unpublished benchmarks

We, our customers, and our partners have run a number of performance tests to benchmark Virsto One against some industry leaders. The bar chart at the top of this post shows the results of one test. Even though the axes aren’t labeled for fear of getting on the wrong side of any licensing police, I will say this: The magnitude of the bars is precisely accurate. Yes, the golden bar shows the performance of Hyper-V with Virsto One in a specific test. The gray bar is really how much I/O we got on the same hardware through an alternative solution from a certain vendor. But I’m not naming names.

Virsto CTO guest post on Microsoft virtualization blog

by Mark Davis, CEO

Posted Tuesday, September 14, 2010 in Technology 0 comments

Tags: false assumptions, Hyper-V, Microsoft, performance, Virsto One, virtualization, VM I/O blender

If you don't subscribe to the Microsoft Windows Virtualization Product Group blog on Technet, you should.

You might want to wander over to read a guest post by Virsto CTO Alex Miroshnichenko, in which he waxes lovingly about Microsoft's new Dynamic Memory feature.

Why does a storage guy care about Dynamic Memory? They're more related than you might think. Alex explains why in his post.

Abracadabra! A new release of Virsto One

by Mark Davis, CEO

Posted Monday, September 13, 2010 in News 0 comments

Tags: bloggers & media, Hyper-V, Microsoft, no dupe, Virsto One, virtualization, VMware, Xen

Unfortunately, I live life as a muggle*.

The Magic of Virsto One Arthur C. Clarke's third Law of Prediction states, “any sufficiently advanced technology is indistinguishable from magic.” When we tell customers what our software can do to improve storage management in virtual server environments – for example, when we say we can cut VM storage hardware costs by over 50% while improving total I/O throughput by 100% – they think there is no way we can do this without magic.

Every once in a while (mostly when I wish I am not a muggle) I get the urge to make magic predictions of my own. Today, I predict, you will LOVE the new release of VIrsto One.

What's new?

Increasing Virsto alignment with Microsoft

We are expanding Virsto One support for three Microsoft technologies.

Integration with Microsoft System Center Data Protection Manager 2010 (DPM) – incredibly fast data protection
Support for Microsoft System Center Virtual Machine Manager 2008 (VMM) – faster provisioning speed with Rapid Provisioning
Expanded Microsoft Volume Shadow Copy Service (VSS) – now supports both writer and provider models

The combination of Virsto One and these Microsoft technologies allow customers an unprecedented level of flexibility and management for their virtual environments. Virsto One now seamlessly integrates into DPM Protection Groups, allowing users to backup and restore virtual machines from Virsto One vDisks. Using the Rapid Provisioning option within VMM, Virsto One users can provision virtual machines (VMs) using no-dupe clones, without waiting for a bit-for-bit copy of the .vhd. The writer and provider model ensure that VSS can perform a full server backup to help protect all data required to fully restore the server.

We are committed to the Microsoft virtualization platform and will continue to deliver capabilities that can make your Hyper-V deployment higher performance, lower cost, and simpler to manage.

Joining the System Center Alliance

We're also announcing that we've joined Microsoft's System Center Alliance. This expansion of our partnership with Microsoft on the technology and marketing fronts will help Microsoft and Virsto deliver superior joint solutions for public and private clouds built on Microsoft Hyper-V virtualization.

This expanded relationship is on top of Virsto's inclusion as a charter member of Microsoft's BizSpark One alliance for venture capital-backed startups seen by Microsoft as having the highest potential. We're proud to be working with Microsoft on multiple fronts.

Other cool new features

Version 1.2 of Virsto One has improved ability to visualize disk usage. We've also added an export feature to complement our existing import capability for simple migration of virtual hard disks into and out of Virsto One.

Oh, by the way, v1.2 is about 15% faster overall than our v1.0 from 6 months ago. Normally, a 15% performance boost is a big deal. But when you're already tripling I/O throughput, people don't get so excited about another 15%! Still, our engineers are rightly proud of the code optimizations that are in the latest version. The next version, parts of which are in QA testing already, are significantly faster still, so nobody should think we're finished with the performance enhancements.

Performance, performance, performance

Did someone mention performance? Since long before our very first beta release of Virsto One, we have been focused on pushing the envelope of I/O performance to relieve the bottleneck caused by the VM I/O Blender. The graph below is a sample of how Virsto One can obliterate the VM I/O Blender for Hyper-V.

High performance VM storage

“The answer, my portenting friend, is…”

Carnac the Magnificent

While this graph shows our intrinsic contribution to Hyper-V, it does beg the questions, “Oh Great Carnac, do other hypervisors suffer from the VM I/O Blender? And will Virsto One ever solve this and other VM storage problems for other hypervisors?”

As to the first question, the answer is an emphatic yes. You needn't be clarvoyant to know this. It is easy to show that the VM I/O Blender impacts VMware, Xen, all the hypervisors.

The biggest difference between the virtualization platforms is that one of them – Hyper-V – has a very high performance, hardware independent, and inexpensive solution: Virsto One.

As to the second question, I will make those magic predictions later.

Staying in the here and now, Virsto One v1.2 is not magic. Merely very clever software.

To learn more about the updated Virsto One, check out the official press release or participate in our webinar.

* That's "non-magic folk" for those of you who are not familiar with the adventures of Harry Potter.

Throwing Hardware at VM Performance Problems

by Mark Davis, CEO

Posted Friday, September 10, 2010 in Alternative Solutions 0 comments

Tags: Dell, commodity hardware, EMC, excessive storage spending, false assumptions, pricing, Virsto One

In Virsto's weekly customer status meeting today, this story was told:

An IT architect we'd spoken to mistakenly spent a bunch of money on new storage hardware to solve a VM I/O performance issue. The architect discovered, after spending US$50K, that he could have gotten even better performance if he'd not spent fifty thousand dollars on incremental hardware, but had instead invested a fraction of that on Virsto One software.

Wasted money on unneeded hardware!

Naturally, the architect discovered this after the money had been spent and the new hardware was installed, too late to get their money back. This is not a particularly large IT shop, so $50K is real money to the architect's employer.

D'oh! $50K down the drain

Imagine you're a virtualization user, in this scenario:

You have a VM performance problem. The problem gets worse the more I/O intensive the applications, and the more VMs per host.

(Hmmm, sounds like the old VM I/O Blender problem. Or perhaps the snapshotting or thin provisioning features of your software or hardware have performance limitations that are well known to industry experts.)
You determine the culprit is storage performance. Your storage subsystem can't seem to sustain the number of IOPS (input output operations per second) needed to feed your virtual machines.

(Sometimes this happens when you try to move from physical to virtual servers without upgrading to a new SAN. Oops, didn't you know that virtualization means spending a lot of money on new storage? Come on, everyone knows that the biggest expense in a virtualization project is the new storage hardware required!)

What to do, what to do?

Shiny new hardware! What your storage hardware vendor will say is that your storage channel is underpowered. And wouldn't you know it, they have just the shiny new hardware gadget to solve your problem. Maybe it's a new disk array, or perhaps a few more disk spindles to stuff into the array they sold you 6 months ago. These days, they might even suggest some flash-based disk technology to give you ultimate performance (and give the vendor ultimate revenue – what a coincidence).

There's nothing wrong with new hardware. Hey, in the 1990s, I was product manager for billions of dollars of storage hardware that IT organizations bought. I can love hardware as much as the next person.

But the problem isn't always what your hardware can do. It's whether your virtualization infrastructure can reap all the horsepower your hardware is capable of.

This is where Virsto One comes in. Time and again since we launched Virsto One earlier this year, customers have found that, without changing their hardware at all, they can get dramatic performance improvements.

How dramatic?

Typically, triple** the I/O throughput. That's what I said: Virsto One can triple the number of IOPS you can get per host server, or per disk spindle.

**Our lawyers hasten me to add, your mileage will vary. How might it vary? Well, the improvement may be less, depending on a lot of variables. Then again, in that same meeting this morning, we learned how one customer's throughput didn't triple. It quadrupled.

Without spending a cent on hardware.

We've seen this in sites using high end EMC Clariion arrays. We've seen such results with Dell Equallogic midrange kit. On ultra cheap commodity RAID. On systems with rotating disks, and systems with flash. Whatever your hardware, there's a good chance Virsto One can help you get more out of it.

If your VM storage seems sluggish, perhaps you should give us a shout before ordering up a truckload of new hardware.

Blog: Real Storage for the Virtual World

by Mark Davis, CEO

Posted Thursday, August 05, 2010 in News 0 comments

Tags: bloggers & media, people, virtualization

A great source of information on virtualization is Virtualization Review. To give you rich technical information straight from the expert horses' mouths, VR editor Bruce Hoard has assembled a sharp group of CTOs to blog about their areas of expertise.

One of those CTOs is Virsto's Alex Miroshnichenko. A couple of his blogs have already been posted on Virtualization Review, and more are in the pipeline. They're informative reading, providing deep but not always politically correct insights, with a good joke thrown in from time to time. The title of the blog, Real Storage, is a bit sly given that the topic is virtual storage. But then virtual storage is a real issue.

Virsto No Dupe (third in a series)

by Alex Miroshnichenko, CTO

Posted Tuesday, July 20, 2010 in Technology 0 comments

Tags: dedupe, no dupe, storage sprawl, Virsto One

Now that you've had ample time to memorize my previous post on cloning, let's wrap up this series by comparing deduplication (dedupe) with Virsto no dupe.

First, a refresher of how we defined data deduplication from my first post in this series: dedupe refers to any technology that analyses data content for the purpose of finding redundancies for subsequent reduction of required space in storage or bandwidth for transmission.

Current dedupe products

The marketplace has data deduplication offerings in software and hardware form. And, as the marketing literature tells me, each of these solution offerings solve world peace. Uh oh, I am having déjà vu. Remember the movie War Games? Remember WOPR and its supercomputing goal of predicting war outcomes? At the end (sorry to spoil it if you still haven't seen this movie from 1983) WOPR concludes the only winning move is not to play. WOPR then opts for a nice game of chess. (How’s that for foreshadowing?)

Scanning and analyzing massive amounts of data is a very resource intensive operation. A typical workflow involves scanning every data block to calculate some kind of reliable checksum (fingerprint), comparing that fingerprint with its known data set, performing additional calculations and comparisons in case of matches and so on. This is extremely hard to do with a fast data stream in real time, precisely because the algorithmic execution and required processing time for a given data set depend on its content. Numerous PhD dissertations have been written on the subject (some of them even defended). And of course, vendor companies have been created, bought and sold to pursue the opportunity. Some of these companies have implemented their algorithmic magic in a dedicated custom designed piece of hardware. Many of these custom hardware pieces are fairly expensive storage devices whose application is data backup and archiving, not primary data storage.

With content-aware data dedupe, like with most resource hungry technologies, we have the option (if not the pragmatic requirement) to delay or schedule processing and perform it in the background when extra bandwidth and CPU cycles are available. This is a valid approach used by many for backup and archiving applications.

There are no examples of – please pay attention to the qualifier – inline software dedupe technologies for online transactional data storage which have been applied on any significant scale. The required processing speeds and memory sizes are simply beyond the current state of the art.

Virsto One clones: optimal for VM storage sprawl

We made a strategic decision for dramatically reducing storage consumption in our initial technology deployment. In my second blog post of the series, I went through a broad list of Virsto One clone characteristics and their benefits. The bottom line is, we believe that any new functionality required by virtualized data centers should not come at the expense of degrading any existing features. In particular, the performance of the online storage subsystem should never be sacrificed in order to support the use of data deduplication.

No Dupe Scalable cloning technology purpose-built for VM deployment is ideal because it deals directly with the primary source of data duplication in virtual machines, namely the tragic waste of resources caused by the constant provisioning of new VM disk images that is fundamental in running a virtual infrastructure. And because dedupe techniques are so resource intensive as to be prohibitive for real time use in production VMs, it doesn't make sense to use the dedupe hammer as the primary tool to drive the nail of VM storage sprawl.

It has been noted numerous times, including in comments about this blog series, that cloning technology does not solve the problem of duplicate data introduced into independent data sets post-provisioning. However, the pain of storage sprawl and excruciatingly slow VM storage provisioning in the virtualized world is acute and immediate. Virsto One clones solve these problems today in a highly economical way, and have beneficial characteristics that no other cloning technology possesses.

No single technology is perfect for all occasions, and great products ultimately combine multiple technologies to solve customer problems. The best products are designed from the start to be extensible and future-proof, and you should be assured that the Virsto One architecture has a natural and elegant way to add data deduplication features. Virsto One is a great product today with a future proof design. I encourage you to try it now.

Chess And WOPR if you are still online, I am up for a nice game of chess. As an opening move, in the near future I'll write about hardware versus software approaches to storage virtualization.

Check. Check. Check. Priceless!

by virsto_admin

Posted Tuesday, June 22, 2010 in News 0 comments

Tags: bloggers & media, people

Rock star. Check.

Product star. Check.

Storage Superstar. Check.

Continue winning combination of recognition. Priceless!

Today, our CTO Alex Miroshnichenko (pronounced just like it's spelled) was named a Storage Superstar by the team at ChannelWeb. Congratulations Alex! We are excited for you. And excited for us. Thanks to the ChannelWeb team for recognizing Alex.

There are a few things to know about Alex. More than a few, but this blog needs posting before everything is yesterday's news.

Alex loves technology. He enjoys understanding how things work and more importantly why they work a certain way. Once he understands the mechanics, he is not shy about taking the counter point in saying it should work differently just to stir the discussion.

On the more personal side, Alex loves to fly. If you need to distract him, just get him talking about wind speeds, free falling, fuel management, auto throttle systems and before you know it, you will be sucked into his flying adventure.

And while the Virsto team enjoys teasing him about his stylishness, let there be no mistake we are proud he is our CTO!

Watch this to see Alex in action talking about Virsto One at Microsoft Tech•Ed.
(Thanks to theyakmiester for sharing his video.)

Virsto at House of Blues NOLA

by Mark Davis, CEO

Posted Wednesday, June 09, 2010 in News 0 comments

Tags: humor, Microsoft, people

Virsto at House of Blues NOLA If you haven't tried Virsto's software, you're missing out on all kinds of coolness.

If you weren't at Microsoft's Tech•Ed virtualization party last night at House of Blues in New Orleans, you missed Virsto's Steve Gearing making a crowd-pleasing cameo appearance with the band.

Steve was invited to sit in on lead guitar and vocal for a couple songs, and brought the house down. The crowd went wild.

Wish I had audio or video, because this picture just doesn't capture the energy and the talent in the room. Nice set Steve!

If you're in New Orleans at Tech•Ed, come by the Virsto booth (#2042) to get Steve's autograph. If you like his guitar playing, you'll love how he gives a demo.

Many thanks to our friends in the Microsoft virtualization team for the invitation to their House of Blues party. Great fun all around. Nothing like a bunch of geeks talking about significant Hyper-V deployments in the midst of the food and the drinks and the music!

Virsto and Cloning (second in a series)

by Alex Miroshnichenko, CTO

Posted Thursday, June 03, 2010 in Technology 0 comments

Tags: dedupe, excessive storage spending, no dupe, storage sprawl, Virsto One

In my prior post, we started to discuss clones. What is a clone, anyway?

It depends on the situation and the nature of objects being cloned. For example, if you could Imperfect clones clone the material contents of a metal box you could be considered for a Nobel prize. (Someone call Stockholm; I think we are cloning the contents of metal boxes…)

If you clone a bank account you are likely to be considered for an extended stay at Club Fed – even if the account belonged to Bernie Madoff.

And on the subject of world domination by evil geniuses, Dr. Evil and Mini Me are an example of imperfect clones.

But I digress. Let's return to where we left off in the previous post.

It is important to know what we are about to clone and that there are different kinds of clones.

Virtual Disk Images

The objects we are dealing with are virtual disk images for virtual machines. These VM images (in a format Microsoft calls VHD, or virtual hard disk) contain operating system executables and VM-specific configuration data.

An executable for a modern OS is quite large. A bare bones distribution of Windows Server 2008 R2 weighs in at about 7GB of genuine Microsoft data bits. Of those, probably 6.99..9GB are binary executables and the remaining couple of bits store default configuration information. The executables will never change. The instance-specific data will change but never grow to larger than 10-20% of the VM image size. Install your favorite application suites like Office 2010 or SAP Business Suite and the total size of the immutable part of your VHD can easily reach 20GB.

Virtual Disk Cloning: Essential to VM Provisioning

To deploy a large number of virtual servers (or desktops for that matter – that's a teaser for a future blog topic), the natural process is to create a template OS image (or golden image) which contains all operating system and application executables including patches and updates, then clone that image as many times as needed. “System provisioning” consists of server provisioning (CPU and memory resources) and storage provisioning – the process of creating cloned copies of the golden virtual disk image.

When some vendors say "we provision storage" as in the above paragraph, it is a fancy way of saying "we have to copy lots of bits". In the majority of cases that's what it is: а plain, dumb copy.

Why dumb? Well, think about it. A single golden image may be used to create a huge number of derivative images, hundreds even thousands. So almost all of the time and disk space it takes to copy those images are wasted copying identical data that will never change.

Ever heard of "VM storage sprawl"? That is where the term comes from. You do the math how much time and space you waste in your environment by copying things around.

A Better Way to Clone VMs

Enter Virsto One. Virsto One is designed for the ideal virtual disk provisioning workflow described above, but without the nasty side effects typical of alternatives.

A cloning operation in Virsto One never copies the golden image data. It creates a virtual disk object, an exact clone of the original. The clone looks, feels, and behaves as if it were a full fidelity copy of the golden image. As far as the hypervisor (or anything else) can tell, it looks like Virsto made one of those dumb copies.

Abundant clones But in fact, we didn't. We created a VHD object that is addressable and usable just like any other VHD, but we didn't copy a single bit of data from the golden image to the clone. All the blocks are shared between the two VHDs. So 50% of the disk space that otherwise would have been taken up by the clone has been saved. Make a third and then a fourth clone. That's a 75% space savings. Twenty clones? 95% savings.

Clones in Action

So much for the initial state of the clone. What happens when data is written to either the golden image or to one of the clones?

As a block of disk space that is shared by more than one VHD is being written to, that block can no longer be shared. The VHD that's being written to needs its own private, non shared, unique version of that block. Virsto One allocates private storage space for that VHD's version of the block.

Unchanged blocks remain shared among all derivatives and do not consume any extra storage. Only truly unique data blocks occupy physical disk space.

I should note that Virsto clones, unlike other types of clones, have some pretty powerful characteristics. For example:

Clone creation is practically instantaneous. You ask for a Virsto clone, and you've got it. Right now.
Clones have the same performance signature as golden images. I/O to a Virsto clone is just as fast as I/O would have been to one of those dumb copies. This is true not only just after clone creation, but even after the clone has lived for a long time.
Clones are instantly available on all nodes in a cluster. You can make a clone on server box A and instantly fire it up as a virtual machine on server B.
Clones (and in fact golden images) are always thin provisioned. Virsto VHDs don't occupy disk space until a block is actually written to. This is similar to Microsoft dynamic VHDs, but remember I/O performance to a Virsto VHD is full throttle, unlike other kinds of thin provisioning.
There is no limit to the number of clones that can be made from one VHD. Alternatives have practical limits of tens, maybe hundreds. At Virsto, we think thousands isn't that large a number.
Clones can be cloned. And clones of clones can be cloned. Et cetera. Arbirtarily complex trees of clones can be created, with Virsto One presenting each clone as a simple linear VHD easily consumed and managed like any other standard Microsoft VHD.

By the way, all this is completely transparent, invisible to guest VMs. So apps you run in these VMs are completely unaffected.

Not a Cow

When we were raising initial venture capital for Virsto, when I reached this point in a presentation to an investor, it was not unusual for an associate in a VC firm to proudly say something smart like "So, this is just a copy-on-write scheme, right?"

No CoW No, it is not. Copy-on-write (CoW) is just one of the implementation algorithms which may exhibit some of the desired properties. However, CoW has a lot of issues with performance, scalability, garbage collection, et cetera. These issues make CoW a poor choice for VM disk images.

If I had to express our technique in a simple catchy phrase I would call it "logging with allocate on log flush". Ok, you might find that neither simple nor catchy but it's my best attempt.

Many people find "flush" to have a bad connotation when applied to valuable data, so we usually use the word "destage" instead. But the essence is the same. Virsto One allocates permanent storage at the time of moving data from the log device to its permanent location. (There is a lot to our logging technique that goes beyond the issues we're discussing here, but we'll leave that to future blog posts.)

Perhaps it is best for now to avoid all the details and answer the question by saying, "No, we don't do copy-on-write. We have secret sauce algorithms derived from our combined decades of experience in the storage and virtualization industries and academia, and we've made that secret sauce specifically for the unique needs of virtual data centers."

Whew, We Covered a Lot of Ground

The good news for you is that the Virsto team has done all the heavy lifting. You don't need to know how we work our magic. All you need to know is that we do make it easy to make lots of super high performance, space saving VM clones, without buying a single piece of new storage or server hardware.

You can draw your own conclusions on how well we have implemented these goals. Just download a trial version Virsto One, and let us know what you think. It's only 6.8MB in size and installs in seconds.

If you've read about or perhaps tried dedupe, you may be wondering how Virsto is different, if at all. In the third and final part of this series, I will contrast Virsto One "no dupe" with "dedupe" alternatives.

Best of Tech•Ed finalist

by Mark Davis, CEO

Posted Thursday, May 27, 2010 in News 0 comments

Tags: Microsoft, Virsto One

Virsto is proud to have been selected as a finalist for the prestigious Best of Tech•Ed competition.

Please come see why at Microsoft Tech•Ed 2010 in New Orleans June 7-10. Virsto is in booth 2042. We look forward to talking with you.

And plan to attend the "Hypervisor-based Storage Virtualization" birds of a feather session (BOF 33-IT) on Thursday June 10 in room 354 at the Morial Convention Center. You might win a Microsoft XBox!

Morgan Stanley’s 10th CTO Summit

by Mark Davis, CEO

Posted Sunday, May 23, 2010 in News 0 comments

Tags: customers, startups

Morgan Stanley’s Technology Department, bellwether of tech startup success, has invited Virsto to participate in their 10th Annual CTO Summit June 16-17 in Palo Alto California.

Alex, our CTO, and I look forward to meeting with Morgan Stanley’s technology specialists and sharing views on virtualization storage.

Many of us at Virsto have worked closely over the years with Morgan Stanley’s reputable IT department, and we value their insight. This opportunity is flattering due to Morgan Stanley’s reputation for the discovery and support of innovative tech solutions at their earliest stages. For example, a little startup called VMware preceded Virsto at the CTO Summit.

Virsto and Deduplication (first in a series)

by Alex Miroshnichenko, CTO

Posted Thursday, May 20, 2010 in Technology 4 comments

Tags: dedupe, false assumptions, no dupe

Since our Virsto One announcement we have received a stream of questions regarding our data deduplication solution. This is understandable given the dedupe hype that media teams are pushing. Even the least sophisticated of hardware distributors feel compelled to put a "we support dedupe" message on the front pages of their data sheets; some even add yellow ribbons.

No Dupe Virsto one does dramatically reduce the amount of storage consumed by virtual disks (VHDs in Microsoft parlance). However, Virsto One does not perform data de-duplication as the industry has coined the term. While dedupe is a useful technique in many environments including virtualization, the specific nature of virtualization workflows calls for what we at Virsto call no dupe. More on that later.

Deduplication 101

This post is not the best forum to provide a comprehensive guide on the diverse universe of data deduplication technology. For this discussion, I will call data dedupe any technology that analyses data content for the purpose of finding redundancies for subsequent reduction of required space in storage or bandwidth for transmission. Readers old enough to remember their high school years before MySpace may also remember a similar data reduction technique called compression. There are substantial differences between compression and deduplication but both techniques rely on analyzing data content.

Virsto One does not use data deduplication in the above sense. We do not look inside data blocks as they pass from virtual machines to I/O channels. (We don't even touch the data, but that's a topic for another post.)

Introduction to Virsto's thinking on VM duplication

Virsto One is built on the realization that in a virtual environment the overwhelming majority of virtual disks are not independent objects. Rather, they are derivatives of each other.

This is an important insight, so let's think about it for a few minutes.

Virtual machine images are fundamentally different from disk images in the physical server world. The magic of virtualization is that now a server (or desktop) is just a big file that is formatted to look like a disk (VHD, VMDK, or other format). This paradigm shift is enormously powerful, and enables all kinds of cool things like live migration of virtual machines across physical servers. It also completely changes the workflow of provisioning servers or desktops.

A new virtual machine is normally – in fact, almost always – created (or provisioned) as a copy of another reference VM. Such reference images are referred to as templates or golden images. As an example, a golden image may contain a reference installation of an operating system with certain applications.

Want a new VM that is similar to one that you created earlier? Just copy the image, simple. Remember, since that VM is just a big file that looks like a disk, making that new VM image means making a copy of the original golden image. Lots of people call that cloning, although you should beware that not every vendor means exactly the same thing when they say "clone".

What is a clone anyway?

Clones

In everyday non-technical conversations, when we hear the word "clone", we think of a separate, independent copy of something. The copy is its own entity with a life of its own, but at the moment of creation, the clone is indistinguishable from its antecedent.

Cloning implies repeatability. If you can make one clone, why not two? Ten? A thousand?

An important concept is that each instance is an individual and can change over time. So a year later, those thousand clones may look very similar to each other, but they are no longer identical. Some of the clones may differ in only very small ways, while other may have more noticeable variations.

This commonsense connotation of a clone is exactly the right way to think of VM image clones. Keep this imagery in mind.

The germ of an idea

For fun, you might want to think about how this notion might suggest a vastly different and better storage architecture for VM environments. If you do, you might start to understand why we started Virsto Software a couple years ago.

In the next post, we'll discuss how this notion of cloning relates to the way virtual machines are provisioned and stored, and how Virsto One optimizes the process. In part three of this triptych, we'll compare and contrast dedupe against no dupe.

And the winner of the Virsto Xbox is…

by virsto_admin

Posted Tuesday, May 11, 2010 in News 0 comments

Tags: people

Drum roll please....

We are pleased to announce the winner of the Virsto Xbox... Janssen Jones, Associate Director of Auxiliary Information Technology (AIT) at Indiana University.

When we called Janssen to tell him he won the Xbox, we asked if he would be willing to share some of his experience with Hyper-V. Here is what he said.

Tell us a little about yourself

Janssen: I have been working for the Indiana University since 2002. I have been in my current role for the past four years.

What does your Hyper-V environment look like?

Janssen: We have been active users of Hyper-V since 2008. My department (AIT) is part of TAP (Technology Adoption Program) for both SCVMM and Hyper-V. We currently manage 180 virtual machines across 16 blades. We have a mix of storage in our SAN that includes solid state disks. The overall environment is 90-95% virtualized. It is very easy to see that in a couple of years we will soon be managing 300 virtual machines.

Biggest Hyper-V pain point

Janssen: For our environment we are seeing pain points around troubleshooting intermittent performance problems. These pain points come from not being able to track bad performance to a single VM. This is mostly due to VMs sharing spindles and host servers. For example, a database backup on one server may cause I/O to queue on a totally unrelated file server, just because all of the back-end server and storage resources are pooled. Finding a way to see the entire picture to identify root cause has proven challenging.

How can Virsto One help?

Janssen: I have some ideas, lots of ideas. A couple that come to mind, that I will be including in the scope of some of my tests are:

A golden image for VDI, create 50 (or more) copies of an image and run our kiosks off that single image.
Using Virsto One in place of CSV for simplification.

Advice for those starting Hyper-V projects

Janssen: Here at AIT we took a phased approach, and I would recommend that for anyone starting a Hyper-V project. We devoted a fair amount of time before the project to architect and plan our virtualization roll out. Our environment allowed us the flexibility to tackle the stand alone hosts first, then integrate clustering for high availability in our second phase. Specifically, in regards to clustering make sure and spend the time in understanding what workflows need 24x7 support. Also seek out mentors and communities where Hyper-V and virtualization in general are being discussed, because there really are resources out there ready to help.

Lastly, what game are you looking forward to playing on your shiny new Xbox?

Janssen: Beatles Rock Band, of course.

Thank you Janssen for sharing your experiences about Hyper-V. And our thanks to everyone who stopped by our booth at the Microsoft Management Summit in Las Vegas. We look forward to giving away another Microsoft Xbox at Tech•Ed in June.

New release of Virsto One!

by Alex Miroshnichenko, CTO

Posted Monday, May 10, 2010 in News 0 comments

Tags: storage sprawl, Virsto One

Six weeks after our initial release of Virsto One we welcome version 1.1. We would not call it 1.1 unless we responded to customer feedback and introduced new features, along with some bug fixes.

I would like to highlight some changes to tracking and disk space reporting.

If you are reading this post than you may already be aware that Virsto One supports thinly provisioned storage and block sharing between storage objects. If not, I suggest reading our architectural white paper.

The physical data blocks in a virtual disk may be highly shared. For example after creating a clone with Virsto One, the total number of physical blocks allocated does not change, but the logical space shown is twice the original size, because now we have two objects (the parent and clone) of equal size. Only after writing to the clone are private blocks allocated, and then the number of physical blocks allocated will increase accordingly.

This seems rather trivial and conceptually it is. However, once you create a large number of virtual disks, clones, and clones of clones – where each of these objects has a number of private allocations – the resulting space allocation picture can be rather complex. The good news is this complexity is what Virsto One keeps track of, so users needn't ever be concerned with it.

Because of Virsto's scalable block sharing, there is usually a difference – often quite large – between physical space consumption and logical space used. Allocating a much smaller amount of physical capacity while presenting a vastly larger amount of logical disk space is one of Virsto One's key value propositions. With Virsto One, you'll buy a good deal less storage hardware than you'd otherwise need.

How much disk space is taken by a particular vDisk?

The answer to this seemingly simple query is, "It depends".

It depends on how much sharing of physical data blocks there is between that vDisk and other objects. The true amount could be as little as zero, and as large as the logical size of the vDisk.

Note that the size of the vDisk as reported to any interrogation will be the logical size of the disk, even if the physical space is zero. As far as any system software is concerned, Virsto One vDisks with their highly shared data blocks behave as if there was no sharing at all. Making Virsto vDisks behave exactly like standard VHDs – even though they are much more space efficient and higher performance – is a fundamental design goal of our software.

Physical disk space reporting in Virsto One is fundamentally a system wide concept. Deleting a virtual disk may or may not lead to freeing up of a significant amount of space. By the way, freeing up of space happens as a background operation so that garbage collection doesn't impact online disk performance.

Space reporting improvements in Virsto One v1.1

Our initial release of Virsto One did not do a good job of showing disk space allocation dynamics. In version 1.1 we changed the way vDisk deletion is presented. Now, deleting a vDisk removes it from vDisk name space and its disk blocks are placed into pool called "pending deletion". We also introduced a system wide gauge which shows three essential parameters of physical disk space space:

Allocated space in use
Free space available for immediate allocation
Pending deletion

We present this dynamic gauge in our Hyper-V Mananager MMC snap-in. This information is also available via WMI and PowerShell to simplify integration with your favorite management framework, including homegrown scripts.

If you still have not tried Virsto One, there is no better time than the present.

Virsto One vVHD and standard VHD

by Alex Miroshnichenko, CTO

Posted Sunday, April 25, 2010 in Technology 0 comments

Tags: Hyper-V, Virsto One

There is nothing like real customer exposure for understanding truth.

Since our Virsto One release we have been getting great positive feedback. Including some valuable questions on various aspects of the product. The one area that seems to bring the most questions is the architecture of our filter driver and our VHD model.

Those questions are usually variations of

“Is your filter driver a disk filter or a filesystem filter?” and
“Do the Virsto One VHDs live on NTFS (CSV)?”

Short answers:

Our filter driver is a filesystem driver for reasons which will become clear once the second question is answered.
Virsto One VHD’s do not “live” on NTFS in the same sense that a regular VHD “live” on an NTFS volume.

Long answers:

We have published a white paper, which I strongly encourage reading first, in particular the section on the virtual disk lifecycle. The way we use the VHD concept is slightly different from what people might be used to in a standard Microsoft environment, and this may cause some confusion.

Some reminders: VHD stands for Virtual Hard Disk. VHD was introduced several years ago with the first wave of Microsoft server virtualization products. In short a standard VHD interface allows a regular file to be treated as a hard disk. There is actually little difference between a file and a hard disk at the logical level. Both present a linear address space which can be accessed in a random fashion, and neither impose any structure on data content. There are differences with respect to granularity of the data addressing. A file can be addressed at any byte address and the size of the I/O request could be an arbitrary integer number of bytes. A disk on the other hand has to be addressed on a sector boundary and all I/O requests must be a multiple of sector size. It is trivial to use a file to model a disk behavior. (The reverse is not true: in fact a filesystem is a very complex piece of system software that does exactly that, i.e. implements a file model on top of a disk model. Well known examples of filesystems are NTFS, FAT32, VxFS and so on.)

So you can think of a standard VHD as a wrapper around a standard NTFS file. VHD adds some metadata fields to facilitate the file presentation as a disk, but the actual data is stored in an NTFS file. For complete details please see the official VHD specification freely available from Microsoft. I deliberately do not mention variations like fixed, dynamic and differencing VHDs because these are not important for the purpose of this discussion. What is important is that all standard VHD types store the data in an NTFS file. Therefore one can say, “A VHD lives on an NTFS volume”.

With Virsto One the situation is different. Virsto One separates the name and the content of the objects it manages. The core Virsto One object is a virtual disk (we call it a vDisk). A vDisk's data is stored on one or multiple volumes. The data is stored in chunks with a minimum granularity of 4MB. Because the chunks can be spread across volumes, it does not make sense to say, “A vDisk lives on a particular volume”.

A Virsto VHD is only a namespace object; it is just an entry in a directory. The directory itself is a virtual directory; it is created and managed by the Virsto One filesystem filter driver (remember the first question!) The filter driver maintains the linkage between a Virsto VHD namespace object and the corresponding vDisk containing all actual data. That linkage is established by Virsto One mount operation. When any system process like Hyper-V Manager opens a Virsto VHD, it in fact opens an object in the filter driver and all the I/O requests are routed to the appropriate vDisk.

You can think of Virsto VHD as of a “virtual VHD”, we introduced the term vVHD to distinguish from the “regular” Microsoft VHDs. vVHDs have all the metadata properties of a standard Microsoft fixed VHD, which enable seamless interoperability with the system and management interfaces in Hyper-V and Windows Server 2008 R2.

From a functional point of view, a Virsto vVHD is the same as a Microsoft VHD, and the two may be used interchangeably. Of course, a Virsto vDisk has some uniquely valuable characterists that standard VHDs don't: Huge space savings through unlimited high performance snapshots and clones, much higher performance I/O thoughput, and way faster VHD provisioning time, to name a few.

I hope that this clarification was useful. Please feel free to comment and ask more questions.

In the next post, I will address a set of questions that revolve around our data deduplication architecture.

The Debut of Virsto One

by Mark Davis, CEO

Posted Tuesday, February 16, 2010 in News, Software Business 4 comments

Tags: Virsto One

Two and a half years ago, we started digging feverishly into the storage issues that are unique to virtual datacenters. A lot of sweat and blood has been expended by a dedicated Virsto team since then. I'm happy to say that today, our first product was announced to the world.

Virsto One is the first and only hypervisor based storage virtualization solution built from the ground up for the unique needs of virtual servers.

It's been an interesting couple years, to say the least. When we started Virsto, there was broad agreement about the opportunity for a new kind of storage for virtual servers. But frankly few people had real insight about what the real problems were, what were the technology and market trends that would impact our success, and what a winning solution would have to look like. It's ok that many "experts" didn't know the answers; figuring these things out is what high technology entrepreneurs do. We still don't have all the answers, but Virsto does have some insights that are far ahead of most people's thinking.

Engaging the market

In the run-up to this product launch, we've done a lot of talking. With end user customers. With resellers. With technology and alliance partners. And most recently, with a broad range of technology and market analysts, bloggers, and the press.

When we walk through why we exist, why traditional storage doesn't work well with VMs, and what we do that is so very different – people get it, overwhelmingly. When we show a demo, and folks see how simple Virsto One is to use, despite the very sophisticated things that are going on underneath, we get a lot of "wow!" responses.

It is not an overstatement to say that people are blown away.

Naturally, one or two cynics didn't want to listen to our story and instead lasered in on side topics before understanding what we do. That's ok. We'll have other chances to interact with the skeptics.

Out in the open

Now that we've taken the wrapping off our technology (or at least off the initial product – surely you don't think we're a one trick pony) we're looking forward to interacting with you. Please connect with us on this blog, Twitter, Facebook, YouTube, or LinkedIn. By the way, we're working on a community website that'll be launched soon.

One of my high school sports coaches used to say, "Son, never bask in your glory." In Texas in the late 1970s, football coaches had a charming way of beating lessons into teenagers. So it is tempting for me to immediately say, "Great launch. What's next?" But let's not go there today. Let's enjoy the fruits of a successful launch for a moment. Tomorrow, we'll talk about what is next.

Our thanks to the many people who have believed in us, not least of which are our families and friends who we don't see enough because we're working so hard on Virsto.

Pre-launch coverage

by Mark Davis, CEO

Posted Monday, February 15, 2010 in News 0 comments

Tags: bloggers & media

As previously noted, Virsto is unveiling our first product tomorrow, February 16 2009.

Besides updates to our website, there will also be a good bit of press coverage on the launch.

A few industry publications

– DCIG, Virtualization Review, and Datamation

– have already mentioned us. You might want to take a look.

Though many people in the United States are taking the day off for the President's Day holiday, you can imagine that things are pretty busy here inside Virsto Software global headquarters. We're very excited to finally come out of the cloak of secrecy. Just one more day!

IT professionals talk about Hyper-V

by Mark Davis, CEO

Posted Friday, February 12, 2010 in Technology 0 comments

Tags: Hyper-V, people

As part of Virsto's beta test program, we've had the pleasure of working with a number of Windows IT professionals who know a lot about virtualization in general and Microsoft Windows Hyper-V in particular.

A couple of these customers have blogs that I recommend. These folks have a great deal of hands on expertise.

Roger Johnson, Enterprise System Group Team Lead at Crutchfield, writes at www.hypervizor.net.

Rob McShinsky, Senior Systems Engineer at the Dartmouth Hitchcock Medical Center, has a website at www.virtuallyaware.com.

In the "interesting coincidences" department, I'll note that Crutchfield is based in Charlottesville Virginia, home of the University of Virginia where I nearly ended up going to graduate school, and Dartmouth College in Hanover New Hampshire is where I did earn my masters degree. Two of the most idyllic places in the United States.

By the way, in a December blog post I promised we'd be launching our first product in early 2010. I wasn't kidding; please check out virsto.com on Tuesday February 16.

New Approaches to Storage Sprawl in IT Business Edge

by Mark Davis, CEO

Posted Thursday, December 31, 2009 in News 0 comments

Tags: excessive storage spending, storage sprawl

Arthur Cole at IT Business Edge posted an article about the problems of storage sprawl in a VM environment. He's right, we think traditional storage approaches are inadquate for the unique needs of virtual servers. We haven't yet briefed IT Business Edge about our technology, but will look forward to speaking with them soon. Thanks Arthur for mentioning Virsto.

Getting ready to launch

by Mark Davis, CEO

Posted Tuesday, December 15, 2009 in News 0 comments

Tags: people, startups

In the summer of 2007, I talked Serge Pashenkov and Alex Miroshnichenko into to quitting their paying jobs and joining me the ranks of the unemployed. It was hugely risky, but there were so many opportunities to help companies with better data storage solutions, and frankly the big vendors weren't (and sadly still aren't) doing much in the way of real innovation. The only way to do anything truly exciting in this industry is to start your own company.

By that fall, we knew exactly what problem to solve: making storage work properly in a virtualized environment. We talked to lots of big and small VMware shops, who all said they loved virtualizing, yet hated how storage worked when they did.

Confident, ready to launch Two years of hard work later, we're about to move Virsto Software into a new phase.

We're transitioning from a development-mode company to a going-to-market company. Early in 2010, we'll launch our first product and take our first customers from beta testing to production deployment.

Igniting the rockets

Virsto has an outstanding technical team. These folks are, in fact, rocket scientists. The power of the technology we've built goes far beyond what will be exposed in our first product, but even this first offering has our beta sites excited to jump, with both feet, into the Virsto solution.

All of us at Virsto Software look forward to firing the rockets on our first product very soon.

Virtual Strategy Magazine: Storage at the front of the 2010 Virtualization Agenda

by Mark Davis, CEO

Posted Wednesday, December 09, 2009 in News 0 comments

Tags: false assumptions, Hyper-V

If one "forecast for 2010" article is good, surely two are better. Here are some very different predictions we had to share with the folks at Virtual Strategy Magazine.

VMblog: 2010 Shines a Light on Mistaken Storage Assumptions

by Mark Davis, CEO

Posted Tuesday, December 08, 2009 in News 0 comments

Tags: false assumptions

It's that time of the year again. As we work ourselves silly trying to finish up year end tasks and prepare for holidays, it is also time for the ritual of predictions for the coming year.

David Marshall at VMblog asked me and others in the virtualization industry for ours. You can see a few thoughts from me here. For a view of other brilliant 2010 prognostications on VMblog.com, go here. I'm sure we're all correct.

You get less of what you tax

by Mark Davis, CEO

Posted Thursday, December 03, 2009 in Software Business 0 comments

Tags: people, pricing

Jerome Wendt makes a good point about storage software pricing on his blog.

He argues that storage software vendors that charge by capacity (terabytes, TB) are shooting themselves in the foot, for three subtly different reasons.

First, customers don’t naturally think of paying for software by the TB. It isn’t how purchasers’ brains are wired. It is hard for vendors to change consumers’ mindset. Even when the vendor is “right” and customers are “wrong”.

Second, TB-based pricing makes the cost of software unpredictable. The customer doesn’t know how much the software is going to cost over its lifetime, because they don’t know how many TB it will be used for. Purchasing and cost accounting people dislike this because they want certainty, up front, on what things will cost.

Third, charging by the TB “taxes” the customer for doing what you want them to do, which is use your cool software on more and more TB. A cardinal lesson from economics is “thou shalt reap less of what thee taxes”. If you want citizens to smoke less, slap a tax on cigarettes.

All three are valid arguments against capacity-based software pricing, but it is the third that most compels me.

Cars, phones, and storage software

A manufacturer of a car that is extremely inexpensive to operate per kilometer might think, “My value proposition is lower cost per distance driven. In fact, the farther customers drive our cars, the better deal they are getting. They should be thrilled to buy our car by the kilometer! So the initial purchase price will be very low, maybe a few hundred dollars, and for that they can drive a short distance. As they need to go farther, we’ll charge their credit card a couple pennies per kilometer.”

If this argument makes sense to you, you’ve probably been part of a storage software pricing meeting. Here’s the line of logic:

Our software helps customers get more value out of their storage.
The more storage they have, the more value customers get from our software.
We want to make it easy for customers to buy our software whether they have only 1 TB or 1,000 TB. That is, we want a very low entry price, but in order for us to make money, we need to get more revenue when people use our software more.
Therefore, we’ll charge by the terabyte.

mobile phone As silly as the car pricing example may sound to you, it has a certain amount of logic. In fact, that pricing model is how many wireless phone contracts work, so don’t laugh too hard.

Perhaps there are lessons to be learned from the mobile phone business. Why is it that per-minute phone pricing works well for some customers, but other customers dislike it?

Hardware versus software pricing

When you sell hardware with your software, or customers can purchase software only from the hardware vendor, pricing is not that difficult. I know because I've done it for billions of dollars of storage hardware and software.

If a vendor's primary driver of revenue is hardware, two phenomena enter the picture. First, making money on selling software becomes a little less important, because they'll make money on the hardware (and services) sale. So the vendor doesn't feel so bad not charging, or charging less, for the software.

Second, it becomes easy to implicitly charge for the software as a part of the hardware price. This is what economists call “bundling”. If one could deconstruct the profit and loss statement of a hardware vendor, it would be easy to see this. There are reasons why a disk drive purchased from your favorite bulge-bracket storage array vendor costs 5X what it would cost to buy the same drive elsewhere. One reason is that you are implicitly paying for the software features of the vendor’s array into which you’ll be stuffing the drive. It most certainly isn’t because the drive is, by itself, 5 times better.

Because of this second phenomenon, software pricing from hardware vendors is, in a real sense, capacity-based pricing. Buy more storage capacity, and you’re paying more for software. But that price is only implicit, which perhaps is one reason why consumers are fine with it.

Please share your input

A pure software business model, however, has no such opportunities for bundling (also known by cynical economists as price obfuscation). This is one of the reasons why it is always harder to price software than hardware in the storage business.

One of the things we’re thinking about at Virsto is how to price our initial product, which is currently in customer beta test. It’s an interesting question indeed. Your input is encouraged.

Evolution of a startup website

by Mark Davis, CEO

Posted Thursday, October 29, 2009 in News 2 comments

Tags: startups

If you’re among the 43.6 million people who've hit www.virsto.com since about 5:29am US pacific time on Monday, you’ll note a couple things.

First, the site has a very different look. Dark and mysterious has been replaced with light and open. We even have graphics to help explain the problems we’re working on. Woot!

Second, there is a lot more content. Double woot!

Third, you’ll perceive a tone, a style meant to foster dialog with IT professionals, consultants, bloggers, integrators, investors, partners, and journalists.

We hope the content, the layout, and all the places where we encourage interaction make you want to connect with us. We’d love to hear from you about anything on the site, its design, or its functionality.

A brief history of virsto.com

One way to trace the state of a startup is by the maturity of their website.

Phase 1: Under construction

Under construction When it came time to name this company, we were fortunate to find a great top level domain that was available. So I got out my credit card, plunked down some exorbitant fee to Network Solutions (what sleazebags – if you used their whois service to look up domain availability, their front running policy meant you had to buy it from them) and I owned virsto.com.

Lame, lame placeholder Phase 2: Exceedingly lame placeholder
With no small amount of effort, I managed to get Google Page Creator (yes, Google was right to kill it) to build this beautiful home page. Yes, that's really what our website looked like for a few months.

Phase 3: Super stealth

We took in some seed venture capital, and were starting to attract attention (we won a couple awards at Under the Radar), so we needed some kind of website, no matter how minimal. Decent looking, if non-descript. Extremely vague about what our company was all about.

Phase 4: Mysterious

A year later, we were pretty far along in product development (our core engine had been running on multiple hypervisors for months), and we needed to accelerate recruitment of employees, customers, and partners. We had to start talking about what we working on. We put up a few pages that hinted at the problems we were working on, masked in an eerie dark background, complete server racks that glowed like alien space ships. That site lasted a few months, up until Monday 5:29am.

Phase 5: Revelation

The full stealth cloak is dropped, and we’re ready to talk in a good deal of detail about the issues we’ve been passionately attacking for the past couple years. We aren’t ready for full-on product launch, but we’re getting close. We hope that our new site is informative, thought provoking, engaging, and even entertaining. Your feedback is encouraged.

Thoughts on the Dumb Disk Fallacy

by Mark Davis, CEO

Posted Thursday, October 15, 2009 in News 0 comments

Tags: false assumptions

Stephen Foskett's highly recommended blog has begun a series on Storage Truths, in which he talks about sacred cows and urban myths of the data storage industry. I've read his thoughtful comments on this industry for a long time, so am looking forward to the series.

His first post on the Dumb Disk Fallacy is on point. But I can't agree with the implication that:

There's a lot more to a high end storage systems than raw disk drives.
All the extra capabilities of enterprise storage systems create a lot of incremental value.
Therefore people should stop talking about the high cost of enterprise storage, and stop complaining about the fact that the cost of some storage systems are many times the cost of the raw materials.

I know Stephen didn't exactly enunciate this syllogism. In fact, he didn't say that at all. I took some liberties. Sorry.

But in between the lines of what he did write, people may infer this line of argument. Even if Stephen didn't make this argument, others certainly do. I agree with points 1 and 2 above 100%. But the concluding point, I couldn't agree with (had Stephen said it).

Another way to look at it

Nobody who knows what they are talking about thinks enterprise storage should cost the same as a raw disk drive from your local low cost electronics bazaar (in Silicon Valley, that's someplace like Fry's). But that argument's a straw man.

The more astute question is why servers are so much more commoditized, competitive, and inexpensive than storage arrays. Do you know of a major server manufacturer that gets the kinds of margins on their gear that EMC or NetApp make on theirs?

Analyze a current enterprise storage P&L and compare it with DEC's or Data General's from a couple decades ago. The structural similarities are not coincidental. Why?

Mainframes and minicomputers

Because the proprietary minicomputer makers built their systems in a way that didn't allow introduction of innovative or lower cost 3rd party hardware or software. That kept prices high and innovation relatively low. The argument from minicomputer makers was that the only way to assure high quality and performance of servers was through vendors' vertical integration, proprietary control, and not coincidentally, high gross profit margins.

For heaven's sake, one couldn't possibly make scalable and reliable server infrastructure from cheap commodity parts you could buy at Fry's. Right?

Most people believed that back then. Does anyone believe that today?

Today, proprietary storage vendors make the argument that storage is so much harder than servers, so the principles that apply to the modern server industry don't apply to storage. Right.

That is a contention I believe can't stand up to scrutiny. And to any degree that it can, it is becoming less so by the day.

An alternate fallacy

Like all good urban legends, the Dumb Disk Fallacy has a grain of plausibility but doesn't hold up to scrutiny.

On the other end of the spectrum is what I'll cheekily call the Brilliant Array Fallacy, embodied in the 3 part syllogism at the top of this post. In many circles it's accepted wisdom.

Just like the wisdom of mainframe and minicomputer economics, back in the day.

Who wants to set the world on fire?

by Mark Davis, CEO

Posted Thursday, September 24, 2009 in News 0 comments

Tags: people, startups

Virsto is in need of stellar product management talent. We're getting very close to putting our first product in our first customer's hands, and we need some to guide our product strategy and roadmap execution.

We have a fantastic platform that can do a lot of exciting things with virtual server storage that no other technology can. We need someone who can guide the market segmentation, definition of product features to meet specific use cases, and product/partner roadmap. We can't go after every opportunity at once, so sequencing the roadmap and building the right bridges to the right partners is a strategic role.

At the same time, there afre important tactical activities like working with our developers on delivery; interacting with customers and partners; managing beta programs; making sure the product fit and finish is perfect; and positioning our products through website, collateral, and interactions with analysts, press, and bloggers.

We need someone who knows server virtualization, storage, and storage virtualization. The technologies, the players, the use cases, the history, the customers. Particularly, we want someone who understands the issues from the perspective of server virtualization. That is, we prefer a person who knows everything about virtualization and a lot about storage, as opposed to the inverse.

Such a person could have managed products at well known companies such as VMware, Microsoft, EMC, EqualLogic, Symantec, NetApp, Citrix, etc. Maybe they have also done an innovative startup or two. The position is located in Silicon Valley. Level is negotiable depending on seniority; we'll match the title to the person. This role will have board-level visibility.

Virsto is a fantastic opportunity for a product genius who is ready to accelerate their career and is capable of playing a leading role in a well funded early stage company that's delivering breakthrough technology into a hot market space.

Do you know someone who walks on water and fits this opportunity? Email us. (No recruiters, please.)

Will storage go the way of the server?

by Mark Davis, CEO

Posted Friday, September 11, 2009 in News 0 comments

Tags: false assumptions, virtualization

If you're in the virtualization business, or you are considering how virtualization will impact your future, you need to study “The Future of Storage: How cloud computing, virtualization, storage efficiency, flash solid state disk and the rise of scale-out architectures change the economics of managing data growth”.

I'll warn you, it's >100 pages. Unlike far too many 100 page analyst reports, it isn't the same 20 pages refactored 5 times, so you really do want to look at it all. But it will be worth your time.

A number of the ideas that drove us to create Virsto Software a couple years ago are discussed in this report.

This report comes from TechAlpha's Juergen Urbanski and George Gilbert. I'd met George an eon ago when he was a star equity analyst at Credit Suisse First Boston, which was banking my software company's IPO. Juergen's experience at McKinsey and NetApp put him in a great position to develop insights into the market where Virsto operates.

Most analysts analyze “what is”, or when they're prognosticating, “what will be soon”. They have a remarkably static worldview, where projections are simple extrapolations. For well-established industries, this is generally a reasonable way to think.

But in an industry that is predicated on innovation, that mode of thinking will always kill you. Especially in a segment like virtualization that is still in its infancy. Believing that the past predicts the future is a sure fire way to be spectacularly wrong.

In the conservative storage industry, almost all “forecasting” is done by grizzled veterans whose eyes are fixated on the rear view mirror.

Not long ago most experts thought that deduplication would never catch on, that the big guys like EMC and NetApp who make a whole lot of money from data duplication would never allow dedupe to succeed as a technology or for Data Domain to succeed as a company. Boy, did the recent multi-billion dollar bidding war between EMC and NetApp for Data Domain disprove that theory!

There are plenty of other examples. 10-ish years ago, Steve Jobs was considered washed up and a poor businessman. Scott McNealy was the smartest guy in high tech. Today, fanbois genuflect at the whisper of Jobs' name, and McNealy's written off as a latter-day Ken Olsen, a man famous for having built, then led to ignominy, one of the greatest computer companies in history.

What's my point? That to see the future requires real analysis, and more importantly, real vision. That it is foolish to envision the future of storage and virtual computing by trivial extrapolation from the present. That Juergen and George are smart analytical cookies with vision.

Disclosure: I spoke with Juergen and George while they were preparing this report, and Virsto Software is mentioned in the document. Except for having paid TechAlpha $99 for a copy of a previous report ("Ripple Effects From Virtualization", also highly recommended), we're not a TechAlpha customer. And while I like their stuff, I don't agree with all of it.

Movin’ on up, to the east side

by Mark Davis, CEO

Posted Thursday, September 03, 2009 in News 5 comments

Tags: humor, startups

…of highway 101. (You had to be living in the US in the late 1970s or early 1980s to appreciate the title of this email.)

Our first real office – after having worked out of my dining room, Serge's garage, and (briefly) Canaan Partners' office – was in Mountain View CA across the freeway from historic Moffett Field dirigible hangar and walking distance to the Veritas/Symantec campus. It was a great home for Virsto in our post-seed, pre-series A phase, when a tiny cadre built our core technology. Collegial, efficient workspace, appropriately shabby and quirky: the bathrooms were outside, and our lab was an old telephone closet with insufficient power and air conditioning. We worked closely together and accomplished an amazing amount of progress.

However, by June we were bursting at the seams. With plans to grow a good bit more, we had to get more space.

Real estate market strangeness

Unlike previous recessions, Silicon Valley has been hit harder than the rest of the country. You can still see offices sitting empty since the .com bust, yet the last mini-boom caused developers to get their shovels out and build even more. There seems to be more available office space within a 40 mile radius of where I sit than in all of Texas.

Despite all the empty space, listed price per square foot hasn't come down much. Arguably the best indicator of future space absorption around here is VC funding, and heaven knows that's not a good story these days – Virsto's significant up round of financing a couple months ago is a rarity. Despite the huge imbalance between supply and demand, most landlords would rather list a space at $2 per square foot per month even though that means a 95% chance of sitting empty for the next year, rather than price it at $1 and get someone in it paying rent. Some of the reasons for this are understandable, some are just economic suicide. But hey, plenty of that happens in high tech (note my alma mater Sun Microsystems, RIP whether the Oracle deal clears or not) so why shouldn't the real estate market be allowed to shoot themselves in the foot if not the head?

Fortunately, we chanced upon a landlord a short walk from NetApp in nearby Sunnyvale with their economic hats on straight, and they wanted the space filled at a price that is significantly above their marginal cost, but way below the artificially-high market asking price. So we snagged it before someone else did. We're fortunate to have the space, and our landlord is lucky to have us as tenants.

The new home

219 Moffett Park Drive, Sunnyvale CA 94089 The space is big enough to last a couple years, for a great price. It's a good bit nicer than our old place. Fully furnished, larger cubes, huge lab with plenty of power & cooling, ample conference rooms, a small kitchen, etc. The team works incredibly hard, and the least we can do is provide a reasonably comfortable place to work all those hours (although the air conditioning has been out today and boy does it get hot in here on a warm summer day).

This is not class A space by any definition. There is nothing impressive or luxurious about it. But it feels like the quality and size of place we shouldn't be in until after our first million dollars of sales.

I fear what Steve Blank calls the curse of the new building. For that reason I was biased towards staying in the old building. It was more run down, felt more “right” for a pre-revenue startup. I don't want anyone in my company to think that banking millions of VC dollars means we've arrived. Economically, the move was a no-brainer because it is very low cost, fully furnished, gives us growth room, and is presentable to partners and customers. The next best alternative was 60% less space at a higher cost, and we would have had to buy furniture then move once again in a year. Only an idiot would have passed up this deal.

Yet the place still feels too good for a company that hasn't shipped a product. Fortunately, our team is inherently frugal, and culturally we keep our noses to the grindstone. Just so everyone remembers to keep that scrappy mode of thinking, we moved the same way fresh college graduates move out of the dorm and into their first apartment. I rented a U-Haul truck, asked the employees to take the morning off from being software developers and become manual laborers instead, and everyone pitched in to do the move to our new office 2 miles away. Over the weekend, the CEO (that's me) and his teenage sons cleaned the 9900 square feet of dirty carpeting. The whole move cost a few hundred dollars. As it should be.

Having said all that, there are two messages I want to leave you with.

First, we're hiring! We've got a fantastic technology platform, a huge market opportunity, and a great team. If you're really good and you know a lot about virtualization and storage and how they're all screwed up, you should be working here. We've now got plenty of space for you.

Second, we're subleasing! If you're an early stage startup (not in a competitive space) and need some really inexpensive, nice, move-in-ready space to park for a few months up to a year, we've got the best real estate deal in Silicon Valley just for you.

Yesterday’s most surprising news story out of VMworld

by Mark Davis, CEO

Posted Tuesday, September 01, 2009 in News 0 comments

Tags: EMC, humor, VMware

Here's the lede from a Computerworld story:

EMC will resell virtualization management software from VMware alongside its own Ionix management platform in an expanded partnership aimed at helping enterprises build private clouds.

Expanded partnership? EMC freaking owns VMware. VMware isn't EMC's partner, they are EMC's chattel.

The article goes on to say that VMware was spun off from EMC in 2007. Umm, not exactly. An IPO of a small minority share, while retaining complete voting control of the company and full control of the board and management is not a spin off.

In other news, GE announced an expanded strategic alliance with NBC, and General Motors and Chevrolet are in negotiations for joint go-to-market plans.

Elephant in the DRS room

by Alex Miroshnichenko, CTO

Posted Friday, July 10, 2009 in Technology 6 comments

Tags: excessive storage spending, VMware

This post was triggered by the latest VMWare article on DRS performance. I have been thinking and talking about this subject ever since I first saw a DRS demo at VMworld in September 2008 in Las Vegas.

DRS (Dynamic Resource Scheduler) is an awesome piece of technology. Dynamic load balancing in distributed systems is a non-trivial task and I applaud the DRS team. A lot of the discussion about DRS is centered on the scalability issues: things like how many physical servers and how many guest VMs (virtual machines) can be managed under DRS. However I would like to draw the attention to the elephant in the room nobody seems to be noticing. Elephant, thy name is storage.

Let's take a closer look at the configuration described in the article. (I am taking only of the vSphere cluster with the servers and storage, not the load generators and other supporting elements):

ESX Hosts (4) : HP DL380 4 Dual socket, quad core Intel Xeon 5450 3.0GHz; 32GB memory; dual port QLogic QLE2462 HBA.

After 15 minutes on the HP website I could not get at the exact configuration described here so I had to spend about 45 minutes in instant chat with HP sales and they came up with a quote of $6,691 discounted to $5,600 if I buy four of them on my credit card immediately. For the sake of simplicity let's assume $6,000 per server including the Fibre Channel (FC) controllers.

Think about it: $6,000 buys vast amounts of computing power today. What's even more important is that I can get exactly the same power from more than one source. I am sure that my Dell sales rep would be happy to beat the HP quote by offering an equivalently configured Dell server for 10-15% cheaper. And I know that the Dell servers will run exactly the same system and application software, I can in fact mix and match Dell, HP and my favorite whitebox servers in the same configuration and DRS will not (and should not!) know the difference.

This is a dramatic achievement of the last 10-15 years - data center servers are true commodities and the prices reflect that.

Now let's look at the storage used in this benchmark:

Storage (1) : CX 4-960 with 188 15K rpm FC disks.

First, notice that the disk size is not specified, so let's make some educated guesses. Getting a quick quote for a high-end storage system is normally a much more involved exercise, so I am still waiting for few sales reps to get back to me. Almighty Google found a reference point at the Reliant Technology website: a Clariion system with 48 15K 300GB FC drives for the list price of $157,500. So I figure to get 188 specified in the DRS article would take 4 of these, yielding a total price tag of more than $600,000! Now, I understand that hardly anybody pays list prices for such systems - but more than half a million dollars!? No wonder the DRS benchmark is listed as a joint EMC/VMware project. It would be very hard – even for VMware – to implement this project as a pure research exercise.

Something is seriously wrong with this picture! Here we have four enormously powerful servers at $6,000 each ($24,000 total), connected to more than half a million dollars worth of disks.

Does anybody building data center infrastructure think it makes any sense to spend 25 times as much on storage as on servers?

The test cases are configured to reach specific levels of CPUload and as far as I know DRS uses CPU and mеmory consumption as inputs to its policy. We are obviously not dealing with a a disk bound configuration. We have a few inexpensive commodity servers connected to a MAINFRAME style storage system.

So DRS has demonstrated that it can achieve "15 - 47% gains in aggregate performance" in a CPU bound configuration when $24K worth of servers are driving $600K worth of disks. Is this interesting technology? Yes. Does it make economic sense? I'm not so sure.

What if you could reduce storage costs by a mere 10%, which is trivially easy to do? With the money saved, you could buy 10 additional servers for an additional 250% compute performance. In that config, you don't need DRS because you're so over-provisioned on server power, and I'd bet your apps will run even faster than the benchmark.

Now what if you could cut the storage costs by 50%, which is also trivial to accomplish? With the money saved, you could add a headcount to IT (or save your own job!) with lots of money left over for a spectacular offsite party for the whole team.

Before anyone complains that I'm being unfair here, let me acknowledge that my numbers are rough. But I could be off by a lot and the point is still valid. It might also be argued that the storage configuration here was a "benchmark special" and would never be used in a real environment. To which I would respond, uh yeah, that's correct. But then we're begging the question about what the whole point of the benchmark was. It's safe to assume that there was a good reason why VMware/EMC configured that many expensive spindles for the benchmark.

Here lies the problem. DRS - or manual load balancing for that matter - can be effective only if you can reliably predict the performance impact of migrating virtual machines from one physical server to another. VMware has done excellent job in providing the tools for this task for cases where the performance is determined ONLY by CPU and memory resources. All the load balancing examples you read about are valid only when the storage I/O channel is grossly over-provisioned.

Note, it is not about the storage capacity, it is about the channel capacity - both in terms of bandwidth and latency. I suspect that in order to achieve the same level of performance described in the article in a carefully managed static configuration, one would need less than half of the total I/O channel capacity (and less than half of the total storage cost). But there is no way to do such reduction in a dynamic configuration, so we have to over-provision.

To be fair, VMware is the absolute leader among all the the virtualization vendors. DRS is the state of the art today, and no other vendor can come close. DRS limitations come from the limitations of the existing storage stack in the virtual environment.

The only way to provide predictable performance and quality of service in virtual servers today is to throw too much money into storage hardware - either in terms of large number of expensive high performance spindles, or a complex (i.e. expensive) array with lots of internal processing power such as a V-MAX class device. In terms of total cost the results would be very similar – your storage is about an order of magnitude more expensive than your servers.

It shouldn't have to be that way. It is time for storage for virtual servers to go the way of the servers themselves – from proprietary mainframes to commodity. The key to that transition is the storage software specifically designed for virtual servers. Then VMWare will be able to repeat the DRS benchmark on a system where costs of storage and servers are much more balanced.

Legacy software

by Alex Miroshnichenko, CTO

Posted Saturday, July 04, 2009 in News 2 comments

Tags: false assumptions, Symantec

As a founder of Virsto, part of my job is to be absolutely convinced that problems we've set out to solve are fundamental, difficult to solve, and important to large numbers of customers. It is very satisfying to find others sharing my views. I recently came across a series of posts by Jerome Wendt on storage management in Microsoft Hyper-V virtual servers.

Jerome correctly identifies some of the problems customer face when configuring Hyper-V installations for real life scalable usage in a data center. The fact is that Hyper-V storage management functionality is scarce at best. In the upcoming 2008 R2 release, Microsoft plugged some of the most gaping holes by introducing clustered shared volumes (CSVs), but Hyper-V still remains woefully inadequate when it comes to storage management.

Jerome correctly points out that life could be a lot easier if customers install Symantec Storage Foundation (SF) for Windows in the Hyper-V parent partition. Specifically, he talks about the I/O path balancing and redundancy features of the SF and superior utilization of thinly provisioned volumes.

I spent a lot of exciting years at VERITAS Software where I was fortunate to design and lead the development of major components of what is now known as Storage Foundation. The concept of introducing a hypervisor-resident storage management layer seems to me (and my other ex-VERITAS colleagues) very natural. So with a certain level of surprise I read that "...In this particular case, Symantec with the assistance of its customers in EMEA, accidentally (emphasis mine) discovered the breakthrough that will make Microsoft Hyper-V a much more viable alternative to the VMware ESX server..."

Accidentally? This discovery was an accident prompted by a customer's use of SF in an unsupported configuration? Is hoping for serendipity what Symantec calls innovation these days? Oh, how far the once-mighty have fallen.

However after some thinking I realized that my surprise was misplaced. How many times since VERITAS was acquired by Symantec have I seen the gross mismanagement of VERITAS technology? Way too many. So I guess I shouldn't have been surprised that Symantec was effectively asleep at the wheel while driving the VERITAS Storage Foundation direction.

One also should not forget the VERITAS SF for Windows was conceived and designed more than a dozen years ago, and has basically been in maintenance mode ever since. (No doubt someone from Symantec is going to object to that statement, but please, the engineers who wrote it know I'm speaking the truth.) x86 virtual servers and their unique storage problems didn't exist back in mid 90s.

Hyper-V customers will need a lot more than a re-purposed progeny of the 90s to bring Hyper-V up to the standard set by VMware vSphere, let alone to solve the hard problems of storage in the new virtual world.

hello, world

by Mark Davis, CEO

Posted Thursday, June 25, 2009 in News 2 comments

Tags: startups

Today, we announced our closing of a $7 million series A venture capital investment, led by Vivek Mehra of August Capital, with continued participation by our seed investor Maha Ibrahim at Canaan Partners. We’re awfully proud to be working with them, and are glad to have their support as we prepare to launch our first product later this year.

It’s a tough market out there for startups raising capital. Absolutely brutal. What does it take for a pre-product IT infrastructure startup to get this strong a funding round from top notch venture capitalists?

An exciting market. The market potential has to be measured in very large numbers (tens of millions of dollars is not interesting), and there has to be reason to believe that the right startup can take a leadership position.
Compelling technology. A technology that’s 20% better than existing products can be fine for a big company’s product strategy. For a startup, it’s woefully inadequate.
Great team. Merely good people don’t cut it. By the way, we’re hiring.
Enthusiastic customers and partners. Many people outside our company have helped us understand how much we can help enterprise and cloud data centers. If your company is interested new solutions for storage in the virtual data center, we’d love to talk with you too.

We’re still pre-product, and we’re believers in the adage that one shouldn’t publicly brag about what their product can do until they can back it up, and we aren’t quite ready for that.

So rather than talking about our product just yet, we want to engage in a dialog about the problems we’ve seen out there. We’re looking forward to discussing the things we’ve seen in the world of storage and virtual servers. It will not be our purpose to throw stones at the competition, but we do intend to describe reality as we see it, and sometimes that will require talking about the inadequacies of existing products.

We’re storage and virtualization people. Experienced, been around the block. Yes, we’re old enough to have learned programming from the writings of Kernighan, hence the "hello, world" reference in the title above. (Although some of us read his work more than a decade after he wrote it. We aren’t that old.)

We think we have some great ideas, but we’re also interested in hearing what you think. So we encourage you to subscribe to and comment on this blog. Stay tuned.