Virsto

I Have a Dream*

Mark Davis — Mon, 11 Feb 2013 20:30:48 +0000

When you start a company, it’s all about the dream, an idea about achieving what most people think impossible, yet inspires great innovation and the willingness of a brilliant team to pour their sweat and blood into changing the world. It has to be that inspiring. Otherwise, why bother?

In the fall of 2007, the founders of Virsto – Alex Miroschnichenko, Serge Pashenkov, and me – had a dream. And today’s announcement that our company is being acquired by VMware is a big step toward making that dream come true.

Here’s how we saw it: VMware (and others) were brilliant when they changed the meaning of the word “server” from referring to hardware, to instead being a software abstraction.

The spark of our inspiration was that someone – and hey, it might as well be us! – had to do the same for storage.

A simple, idealistic idea that everyone in the storage industry knew was crazy. Just as the original idea of VMware was just academic pie in the sky when they started. Cute idea, but it’ll never catch on, all the experts said.

Well, turned out the experts were wrong.

Simple Dream, Complex Problem

Building a storage hypervisor is no small task. This is serious computer science kids. The amount of intellectual capital that has gone into building Virsto’s technology is staggering. And if the R&D problems weren’t enough, the business model challenges of trying to transform the storage industry are equally daunting.

Part of the reason I’m so excited to be combining forces with VMware is how it will help us realize our dream on both fronts.

VMware has some great engineers and technologies that complement what Virsto brings to the table. Together, we’ll build even more amazing solutions.

At the same time, VMware’s unique position in the datacenter universe gives Virsto a powerful way to go to market and catalyze growth of a substantial new segment of the enterprise and cloud storage market.

The Software Defined Datacenter Just Got a Lot Closer

It was the highly aligned way at looking at the opportunity that got VMware and Virsto excited about working together. As we spent time together in recent months, we discovered a remarkable degree of synergy – from high level vision down to minute details of implementation architecture. We spent a lot of time finishing each other’s sentences in those deep dive diligence meetings.

VMware’s vision for the software defined datacenter is far in front of the rest of the industry. Virsto is proud to be able to contribute to VMware’s leadership of the future of enterprise and cloud datacenters.

Exciting Times for the Storage and Virtualization Industry

This acquisition will enable more and more VMware customers to benefit from virtualization technology being optimally applied to storage, with beneficial implications on performance, manageability, and cost. VMware’s solution portfolio is the best in the business to get the most out of compute, storage, and networking resources, driving higher performance, lower cost, and more flexible data center environments that are agnostic when it comes to the underlying hardware. This is the power of VMware’s software defined data center vision, wielded to customer benefit.

We are quite excited to become part of VMware!

*The title of this post is a hat tip to Martin Casado of Nicira, who upon their acquisition by VMware wrote a blog titled “To Infinity And Beyond”, quoting a famous philosopher and societal change agent. So I thought I’d quote one as well.

The Missing Link in Software-Defined Storage

Eric Burgener — Fri, 26 Oct 2012 00:25:53 +0000

At this year’s VMworld 2012, held in San Francisco at the Moscone Center, the “software-defined data center” was a major topic. The idea is appealing: make the entire IT infrastructure completely fluid so that all resources – compute, network, and storage – can be quickly and easily allocated or reclaimed to meet dynamic business requirements (including predictable performance), thereby extending the flexibility and cost savings benefits virtualization provides across the entire IT infrastructure.

If you take what the server hypervisor did for compute resources as the analogy - allowing a single control point to manage the dynamic allocation of CPU and memory resources that are consistently used very efficiently – this part of the puzzle has been "software-defined" and solved. A solution to this problem for the network side is well on its way to being solved - note VMware’s recent acquisition of Nicira, and Oracle’s smaller purchase of Xsigo, as steps towards allowing more of the IT infrastructure to be managed as purely virtual resources directly from the hypervisor.

It’s clear that this same transformation needs to happen with storage. In essence, software-defined storage is the missing link, the last hardware bridge to cross on the path to the software-defined data center.

So just what is “software-defined storage”? Well, if you stick with the earlier analogy we drew with the server hypervisor and what it did for compute resources, storage virtualization is clearly part of it but is not sufficient by itself. Server hypervisor technology not only made compute resources fluid and much easier to manage, it also significantly increased the utilization of those resources to cut the costs of the compute hardware infrastructure. For servers, it was the server consolidation that virtualization enabled that provided this huge boost in utilization, effectively allowing customers to support the same application workloads with much smaller and less costly compute infrastructure.

Storage resources clearly DO have to be virtualized, but the “improved utilization of existing resources” question manifests itself a little differently. If you’ve played with virtual environments, you already know that storage performance suffers in most type 1 hypervisor deployments relative to what you’d expect it to be based on your experience with storage on physical servers. This “VM I/O blender” issue is discussed at length in other blogs on this site (and also by other vendors who have noted it) so I won’t go into it here, but the 30% - 50% storage performance slowdown that virtual administrators experience in these environments touches not only runtime performance, but your ability to use storage capacity optimization technologies like thin provisioning, valuable operational features like snapshots and clones, and quickly provision the storage you need.

You have 15K RPM disks in your environment that should be capable of handling 180 IOPS (on average) but may actually only be delivering (from your guest VMs’ points of view) 30-45 IOPS. This is due in part to the randomness and write-intensivity of most virtual workloads (the VM I/O blender effect) as well as thin provisioning and snapshot/clone implementations. If you could write sequentially to that same 15K RPM device, you’d find that it could deliver more like 2000 IOPS… and that’s on a per spindle basis. How do you unlock that unused potential?

THAT’S the other challenge that you should expect software-defined storage to address. It has to not only virtualize heterogeneous storage resources, but it also has to unlock the performance that is imprisoned within them to allow for the instant provisioning of high performance, space-efficient storage.

Over the last year, the term “storage hypervisor” has been gaining currency and adopted by vendors including IBM, NetApp, HDS, Datacore and Virsto. Definitions of the storage hypervisor show that it can be an attractive way to deliver on software-defined storage. But what constitutes a storage hypervisor?

To truly be the missing link in delivering on the software-defined data center, it has to be implemented in software (so that it integrates with the native hypervisor while preserving those familiar workflows), virtualize heterogeneous storage resources to provide flexibility in allocation, and inherently increase the utilization of existing storage. That means that it has to make your existing storage support instantly provisioned, high performance, space-efficient (read thin provisioned and zero copy clones), VM-centric (NOT LUN centric) storage objects WITHOUT requiring that you purchase higher performance or additional storage to get there. It goes without saying that if you DO buy a higher performance or additional storage, it should speed that up too.

Virsto just recorded a webinar on this topic entitled “Software-Defined Storage: How It Will Improve Your Virtual Environment”. We have a storage hypervisor that runs on both vSphere and Hyper-V and delivers on the requirements above, thereby offering a way for enterprises to take that next step that transforms storage to deliver on the promise of the software-defined data center. You can check that out here.

From a bottom-line financial point of view, what the Virsto Storage Hypervisor will do for you is give you up to 10x the performance (IOPS, throughput) on a per spindle basis that you get with your storage today, while lowering storage latencies, allowing you to cut storage costs by generally around half while meeting your EXISTING performance requirements with your EXISTING storage (or any new block-based storage if you’re embarking on a new rather than an existing project). That storage is space-efficient, can be provisioned almost instantly (so you can spin up the high performance storage you need as fast as you can spin up a new VM), and looks like a native storage object (VMDK on vSphere, VHD on Hyper-V) so that it is managed with familiar native hypervisor workflows while delivering these benefits.

If the software-defined data center idea appeals to you, Virsto will help you get there. If you are working your way toward that vision, Virsto will cut your storage costs by 50% and deliver highly scalable snapshots and clones for VDI, snap reverts, test and dev, rapid database clone creation, and backup.

Virsto “Storage in Virtualized Environments” Haiku Contest: The Winners Are…

Gregg Holzrichter — Fri, 31 Aug 2012 16:00:28 +0000

After an exciting month of poetry-meets-storage in virtualized environments, we are happy to announce the winners of our first ever Haiku contest. The contest pulled in almost 100 creative haikus and we are pleased to have decided on a winner of the iPad 3 grand prize as well as the second and third place cash prize winners.

The Twitter-based contest asked contestants to share a haiku poem via Twitter expressing their best or worst storage experiences in virtualized environments in a fun way. From the hilarious to the uncanny, we saw a little bit of everything. The judging panel had a difficult (but fun) time choosing a winner.

…AND THE WINNERS ARE:

Grand Prize (iPad 3
@Kaizeneer
Agile Storage math

Virsto software plus VSphere

Equals ten times more!

2^nd Place- $250
@SQbenedict
Without You, Virsto

I'm Virtually Nothing

You Vir-Stole My Heart...

3^rd Place- $100

@question10
More capacity

Delivers ten times as much

Virsto is the one.

Winners were selected based on creativity, humor, and relevance and announced via Twitter. Judges also made sure all submissions adhered to the contest rules and guidelines.

Thanks to all of the entrants and our judging panel for a great contest!

Virsto for vSphere 1.5 Optimizes VM Storage for the Software-Defined Datacenter

Eric Burgener — Tue, 21 Aug 2012 02:01:30 +0000

We’re pleased today to announce the next generation of our storage hypervisor, purpose built software designed to deliver better performance, capacity utilization, and storage agility for virtual machines.

The news:

Virsto for vSphere 1.5 now Generally Available
VM-level storage management and seamless integration with VMware vCenter and View Manager
Virsto’s storage hypervisor provides the most efficient provisioning, management and utilization of storage for the Software-Defined Datacenter

We hoped at VMworld last year when we announced our intention to support the VMware market that the demand for a solution built from the ground up to solve the challenges that virtualized workloads present to storage would be strong. Having just delivered 500% year / year revenue growth and dozens of engagements spanning every type of virtualized workload – from VDI, to test & dev, to database virtualization, we’re excited to be solving a huge industry pain point with a truly unique approach.

Virsto has always given you a way to increase your storage IOPS without having to buy a single new piece of hardware. But with our 1.5 release, we’ve given you that plus created out of the box integration with VMware View that is seamless. If you know how to manage View desktops with View Composer and View Connection Server without Virsto installed, then you know how to manage them with Virsto installed. Just add software and enjoy 10x better performance with 90% lower capacity consumption.

Standard VDI workflows have always had a few little quirks. When you provision desktops, there is a failure rate for desktop creation. With native tools, you specify a certain number of desktops, and most of them will be successfully created. There are some that don’t, and you have to find those, delete them, and create the additional ones you need – manually. Virsto does that for you – if you ask for 1000 desktops, you get 1000 desktops. We handle any issues associated with desktop creation failures underneath the covers so you don’t have to.

We’ve also got a “fast prep” option that does the same thing for you that the “quick prep” option that View offers, except of course you’re getting high performance, space-efficient, cluster-aware storage that provisions almost instantly. And our “fast prep” works with View to cut the desktop deployment times pretty significantly regardless of whether it is during the compose, re-compose, or refresh operations.

Ever had a case where you’re installing a Windows patch for your desktops, and part of the way through the re-compose operation you get a failure? Now you’ve got a case where you’ve got some desktops in a pool with the old OS image, and some with the new. Kind of a pain to have to go in manually and resolve this issue with native tools. We’ve implemented a re-deploy recovery capability that handles this issue automatically for you – if you have a failure part way through, you just click “Re-Deploy Recovery” and it identifies what still needs to be done and completes it so you end up with what you originally wanted – all the desktops have the new updated OS image.

So what does this all mean for the VDI administrator?

You can provision desktops at least 75% faster than before, which may mean you can do more refreshes per year (installing the latest updates, etc.) which should provide more reliable, more secure desktop environments
Your storage will be delivering 10x the IOPS and, combined with the fact that Virsto storage is thin provisioned, you need a lot less of it to meet your performance requirements  cost savings
You spend a lot less time doing manual re-work during typical VDI workflow operations, so its easier on you
If you want to back up your desktops, you’ll be using the industry’s fastest, most scalable snapshot/clone implementation for snapshot backups that will be running on whatever hardware you already own

Aside from the operational advantages that Virsto brings to these environments, we will cut your storage cost/desktop by at least half without requiring that you buy any new hardware. For most people that have shelved a VDI project due to cost issues, storage has been the problem. Storage costs generally comprise anywhere from 50% - 85% of the cost of the backend infrastructure. With Virsto, you’ve got a great shot at getting to that magical $200/desktop price point – something that will allow you to enjoy the benefits that VDI offers… affordably!

The Software-Defined Data Center Is Still Missing… Storage

Eric Burgener — Fri, 03 Aug 2012 18:55:29 +0000

You don't have to have a crystal ball to know that there will be some major acquisitions within the next year that will round out the virtualization requirements (CPU, memory, storage, network) for the software-defined data center.

In the wake of VMware’s $1.26B acquisition of Nicira, there’s been a lot of discussion around the concept of the software-defined data center. When VMware introduced the concept of x86-based virtual machines, it was an idea that brought immediate value to IT. Ironically, despite its significant value, it was a woefully incomplete concept.

Software aside, a computer has four basic components: CPU, memory, storage, and networking. Server hypervisors effectively virtualized CPU and memory resources, but didn’t follow through to leverage those same concepts to get the most out of storage and networking resources. Over the last 5 years, there has been a lot of venture capital invested in startups that were looking to extend valuable virtual computing concepts to these two areas as well.

Two recently announced virtual networking acquisitions - Nicira by VMware and Xsigo by Oracle – were clearly targeted at building out these respective vendors' virtual infrastructure stacks. Nicira and Xsigo don’t do exactly the same thing – Nicira’s Network Virtualization Platform (NVP) enables the creation of virtual layer 2 and layer 3 networks that operate independently of the underlying physical network whereas Xsigo is focused more on virtualizing I/O with a top-of-rack solution that reduces cable count – but both are focused on virtualizing networks in a way that lets you better leverage heterogeneous networking hardware in virtual computing environments. The bottom line value proposition for end users with this type of technology is that they can extend the capabilities of networking gear to provide maximum configuration agility the same way server virtualization technology did for virtual machines, and in the process support better business responsiveness along with cost savings.

You don’t have to have a crystal ball to know that there will be some major acquisitions within the next year that will bring technologies that do the same kind of thing for storage into the fold for major hypervisor vendors. Take VMware, for instance. They developed the technology to virtualize CPU and memory in-house, they just bought Nicira to help extend that vision to networking, and there is one major component left that stands to benefit from a similar treatment: storage.

Turning to storage, it’s clear that ALL enterprises are woefully underutilizing storage in virtual computing environments today. You need a lot more of it even than in physical environments to meet any given performance target because of the VM I/O blender problem. [You can see other blogs on our site and other vendors’ sites that discuss this performance degradation problem – you’ve got this big engine but most of its power can’t even be accessed.] But the problem isn’t just about performance – it’s about delivering the most performance out of your storage hardware while at the same time using capacity much more efficiently (storage capacity optimization technologies like thin provisioning, etc.), providing a scalable snapshot/clone implementation that does not impose performance impacts, and providing the same kind of instant provisioning for high performance storage that we have today for virtual machines.

The comment on snapshots/clones may not be intuitive. Because of the severe performance degradations (often as much as 90%) imposed by thin provisioned clones and the scalability limitations (some popular arrays in common enterprise use only support 8 snapshots per source disk), administrators have become conditioned to avoid using them or to use them only sparingly. In the same way that server hypervisors unleashed an ability to get a lot more out of CPU and memory technology, removing the performance and scalability constraints of conventional snapshot/clone technology can enable entirely new ways of using them in virtual environments. The ability to take snapshots at the hypervisor level (and make writable clones from them) becomes a lot more valuable in test/dev, VDI, backup, DR and other environments and/or workflows when you remove these constraints.

Virtual computing has to keep an eye on enterprise requirements as it looks to bring all four areas – CPU, memory, storage, and networking – under the same wing. That means that it has to support data integrity, failover, existing enterprise backup regimens, replication for DR purposes, and of course heterogeneous storage. This means you can’t just slap some SSD into your host and be done with it. There has to be a more intelligent way to address the problem (which may or may not leverage SSD as part of its solution) that takes all these requirements into account.

This all brings me back to the software-defined data center. To truly realize the promise of that concept, you need to appropriately virtualize CPU, memory, storage, and network. VMware, Microsoft, and Citrix know that. CPU and memory are done. Storage and networking are left. The acquisition frenzy is already beginning on the network side. The question is: what to do about storage?

Virsto provides one answer to that solution that meets all the above requirements: the storage hypervisor. The missing link for storage in the Software Defined Datacenter.

Calling All Aspiring IT Poets; Announcing Storage in Virtualized Environments Haiku Twitter Contest

Gregg Holzrichter — Tue, 24 Jul 2012 07:01:13 +0000

Today we launched a storage in virtualized environments haiku poem contest on Twitter. The contest gives anyone an opportunity to share a haiku poem via Twitter expressing their best or worst storage experiences working with virtualized environments.

We all know that server virtualization fundamentally changes the way workloads are presented to storage, with performance penalties, capacity overprovisioning, and increased challenges between server and storage teams for efficient workflow and operations. Exceptional and horrible storage experiences in virtualized environments make lasting impressions and we want to hear all about it. This is an opportunity for you to tap in to your creative side for a chance to win an iPad 3 or cash prize.

A panel of judges will select the first, second and third prize winners based on creativity, humor and relevance. Winners will be announced via Twitter and at VMworld 2012 on August 28, 2012.

To enter, write and submit a haiku poem (17 syllables total with three lines of 5-7-5 syllables) on your experience. Each line of the haiku should be separated by a comma or slash. All entries must be submitted via Twitter to @Virsto, and include the hash tag #VirstoHaiku. All entrants MUST be following @Virsto in order for your haiku to qualify.

For official rules and prizes, please visit: http://virsto.com/virstohaiku

Video chalk talk: Current storage industry trends

Gregg Holzrichter — Tue, 19 Jun 2012 07:01:41 +0000

Virsto Software's Gregg Holzrichter gives an overview of the latest industry trends in hardware and software-based innovations for solving the problem of storage inefficiency in virtualized workloads.

Understanding How Virsto Saves On Physical Storage Capacity

Eric Burgener — Mon, 18 Jun 2012 17:50:37 +0000

Virsto combines a number of different technologies which simultaneously work together to save 90% or more physical storage capacity relative to the way most people deploy their storage in virtual computing environments. As people learn about our technology for the first time, understanding exactly how we save space may be a bit tricky to follow.

In sizing your storage configuration, you generally take three issues into account - performance, availability, and capacity – and in that order. On the performance side, once you’ve determined your peak and steady state I/O requirements, you can quickly determine how many disk spindles you need to meet those performance requirements for a given disk technology (i.e. FC, SAS, SATA, etc.). Make sure you understand your read vs write ratios as you’re doing this since read/write splits can be 10/90 in VDI environments, and spinning disks (and SSD) don't perform nearly as well for writes as they do for reads.

For availability, you need to ensure that you have sufficient spindles to meet your performance requirement given the overhead associated with the RAID level(s) you have chosen for your storage. Since RAID creates additional I/Os, you’ll need to make sure that you can meet the need for your RAID level(s) and still provide for the IOPS your VMs need.

Finally, you need to make sure you have enough physical capacity to house your data. Usually the spindle count requirement leads you to over-provision from a capacity point of view, but you'll need to check this to make sure. For example, if you have 500 VMs, each of which is using 50GB of storage, you’ll need 25TB of storage. If you planned to meet your performance requirements for 20,000 IOPS with 15K RPM 450GB FC disks, each of which could handle 200 IOPS (this is just an example), then you would have reckoned you need 100 disks. But 100 450GB disks gives you 45TB of space, more than the 25TB you calculated you needed (even after taking your RAID overhead into account).

Don’t forget that, because of the very random, very write-intensive I/O patterns in virtual environments, a disk that is rated by its manufacturer to provide 200 IOPS is not likely to provide anywhere near that in actual usage from your VMs' points of view. This is one of the reasons why it’s so important to run a pilot with your actual workload (or something near to it) so you understand what you are really going to get, and you can run your calculations from there.

This brings us to the first (and probably the biggest) reason that Virsto reduces storage capacity consumption: by sequentializing the I/O stream as it comes out of the host using a log architecture, you will get significantly more IOPS out of each disk. If you needed 100 disks before, you will now need a fraction of that to get the same performance.

Because of this significant reduction, you now need to look at your capacity requirements. With the much smaller number of disks, do you still have enough physical capacity to meet your storage requirements? Let’s say that instead of needing 100 disks, you ended up needing 20 now (this is quite conservative based on what we’ve seen in customer settings, by the way). 20 x 450GB = 9TB, which is quite a bit smaller than your 25TB. Do you have to add disks to get back to that?

That’s where Virsto’s thin provisioning comes in handy. While you may determine that you’d like each VM to have 50GB of space, in actual usage you probably will not need anywhere near that unless you leave a VM spun up for months or even years. When you first create it, you may write as little as 100MB – 500MB of actual data to it. The beauty of thin provisioning is that it only allocates storage as it is actually consumed.

If all your VMs stay up long enough to consume their full 50GB each, then you will need 25TB for this config. But until that happens, you can get away with a lot less storage capacity. Your VMs think they have 50GB each, but Virsto is managing their perception vs their actual usage. Our experience has been that most customers reduce their actual storage capacity consumption (used, not just allocated) by anywhere from 75% - 90% when they move to Virsto.

But what about the major performance hit associated with thin provisioning? Conventional implementations can slow your write performance by anywhere from 75% - 90% on a per spindle basis, driving up your need for more spindles. Not so with Virsto – our log architecture completely insulates the guest VMs from any performance hit associated with thin provisioning.

Don’t forget – you’ll need to take the RAID overhead (both for capacity consumption and additional I/Os) into account when performing these calculations.

Finally, Virsto’s single image management capabilities contribute to significant space savings as well. Anytime you are creating more than one VM from a single VM template, there is the opportunity to use Virsto’s innovative snapshot/clone technology to reduce capacity consumption without any performance hit. Virtual desktop infrastructure (VDI) is a perfect example of this where hundreds or even thousands of VMs are cloned from a single golden master. In this case, Virsto allows you to maintain one logical copy of the golden master (in the form of a Virsto vClone), and handle all the reads to that disk (e.g. the base desktop image that includes the OS and the key applications shared by all the desktops) from that single copy. Virsto’s storage tiering allows these shared blocks to be placed in a very high performance tier to ensure high read performance. Instead of having hundreds or thousands of copies of the same disk, you have one, yet each desktop thinks they have exclusive access to their own clone. This approach is better than de-dup because it never creates redundant blocks in the first place that have to be de-dup’d out.

It's like having a high speed carrot shared, not by two guinea pigs, but by literally thousands.

So there you have it. The amount of physical storage capacity that Virsto saves you is a function of 1) more IOPS per spindle, 2) high performance thin provisioning to eliminate allocated but unused storage, and 3) single image management capabilities.

A Rising Tide Floats All Boats

Eric Burgener — Wed, 30 May 2012 17:02:21 +0000

Virtual infrastructure today suffers from a significant problem: because of its very random, very write-intensive I/O patterns, it uses storage very inefficiently. This is true regardless of whether we’re talking about server or desktop workloads. I just attended the BriForum 2012 event in London where there was a lot of discussion about how to address this problem in desktop virtualization environments, and it really struck me that focusing on it as only a virtual desktop issue is strategically shortsighted.

Just because you’ve had great success virtualizing your server workloads doesn’t mean you will automatically achieve that same success with desktop virtualization. Much of the discussion in the conference sessions was around how important virtualizing for the right reasons is in achieving desktop virtualization success. But if you’ve decided to virtualize your desktops (and there actually are some great reasons to do that, which is the topic of a separate blog), then it seems to make sense that you would leverage, at least in part, the same virtual infrastructure products you’re already familiar with from the server side. The infrastructure components that seem to be the best candidates for this are the hypervisor, the virtual infrastructure management tool set, and the storage infrastructure.

But what got me thinking is this: that virtual server infrastructure has a built-in storage performance problem that you’re probably already aware of from your virtual server deployments, and the I/O patterns in server-hosted virtual desktop environments compound that problem, degrading storage performance even more and driving even higher storage costs to address it. Desktop traffic is even more write intensive than server traffic because of the way Windows interacts with storage, and you have much larger variances between peak and average IOPS requirements due to phenomena like boot, login, application, and logout storms.

There are all sorts of options which vendors suggest can resolve this: buy more disk spindles, upgrade to higher performance storage, use client-side virtualization with local client-side storage, use server-side virtualization with local server-side storage with VDI or streaming, integrate SSD into your environment, use the read caching features now included with most virtual desktop products, deploy various tiered storage configurations, and buy virtual appliances that combine storage technologies like deduplication and write through caching with DRAM-based local caches in the hosts. Most of these approaches require that you buy additional hardware to overcome the performance problems, and they are addressing the core problem (random, write-intensive I/O) in a manner which is conceptually way downstream, where you have to throw more hardware at it to fix it.

You could characterize this approach as follows: virtual servers slow storage down, so you should throw more hardware at the problem. Virtual desktops slow it down even more, so you should throw even more hardware at them. More disks, more SSD, more memory… these hardware add-ons are expensive, ideally you’d like to buy as little of it as possible to meet your performance requirements.

But if we make the way that the hypervisor handles storage I/O more efficient – an approach by the way which lessens the magnitude of the problem for both virtual servers and virtual desktops – we have a lot smaller problem to deal with (if we have one at all). One of the things that potentially makes this approach attractive is that it is a single investment which addresses a pressing and common problem in both server and desktop environments, but if we’re going to evaluate a potential solution for use in both environments, it has to adequately address the core requirements in each area. And the widespread need for failover (or high availability (HA), as some folks call it) in server environments throws a wrench into any “speed up” approach today that relies on local storage, regardless of whether that’s RAM, SSD, or spinning disk. You can’t fail over without losing data in these types of environments (if you can fail over at all).

In my discussions with end users (and chatting with attendees after my session “Storage Challenges in VDI Environments”) at BriForum it became clear that, while Virsto’s high performance, space-efficient, instantly provisionable storage capabilities were interesting, the fact that it could be used across both server and desktop environments (because we fully support the type of failover without data loss that servers require) piqued their interest. Although most attendees either had or were thinking about virtual desktops, every attendee had virtual servers.

In a nutshell, Virsto effectively turns the very random I/O into an almost 100% sequential pattern before it writes it to disk, so what your VMs see is the performance of your storage, whatever it is, operating in streaming, not random, mode. The more random and more write-intensive your environment is, the bigger the increase in IOPS you get out of your existing storage (we don’t sell hardware but if you buy more of it, we’ll speed that up too).

What I got from attendees was this: so your software makes my core virtual infrastructure at least 10x more efficient in terms of how it leverages the storage I already own, and I can use this same software across both the hosts I’m using for virtual servers and the hosts I’m using for virtual desktops? So if my storage is giving me 10x more IOPS than I get out of it now, I may not even need to invest in any of these other desktop-specific hardware approaches (not to mention I need a lot fewer disk spindles), and if I do, then I will need a lot less of it to meet my needs. And I’ll get a huge speed up against my virtual server environments with the same product as well?

Yep.

Thanks, by the way, to Brian Madden and his team for putting on a great show that was packed with practical technical advice on why to implement and how to most effectively manage virtual desktop environments. If you haven’t been to this show before and you’re thinking about desktop virtualization, I would highly recommend it.

Virsto Adding Citrix XenDesktop Support

Eric Burgener — Wed, 09 May 2012 14:29:30 +0000

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.