Moving the needle on VDI

Perhaps you heard that last year, or the year before that, or the year before that…that it was going to be the “year of VDI”. The saying has become a joke among those in the desktop virtualization space. The high expectations of analysts, consultants, or the vendors who pimp their products has never reached the broad adoption (or interest) levels needed to consider VDI mainstream, let alone to call it the “year of VDI”.

When I’m in the company of other Desktop Virtualization consultants conversations often lead to debates about what new features they need, needed enhancements to the remoting protocol, why they think PVS is so much better than MCS…the list goes on and on and on. I used to participate in the lively debate, but no more. What I realized is none of these discussions impact the broad adoption of Desktop Virtualization as an alternative to running an OS and applications decentralized on physical endpoints. While everyone is focused on beating the competition, we’re not moving the Desktop Virtualization industry forward. We’re all focused on the minutia. I say this as someone who spent 5 years agonizing on details to differentiate the solution I was selling from the competition. It is what I did in order to convince customers of my solutions superiority and win the deal, but it didn’t change the reality that Desktop Virtualization was a niche solution for 1% of the world’s users.

There are 2 major barriers to Desktop Virtualization, cost and end-user experience. Many of the issues around end-user experience continue to improve but we’re never going to reach nirvana, the big problems are Microsoft problems and those solutions usually involve moving the problem to a new place. The problems around app deployment, app virtualization, app streaming, profile management, etc…those aren’t going away any time soon and although there are some emerging app layering technologies they aren’t without their issues or risk either.

However, one area we have control is cost. Cost has many components but for years the biggest one has been storage. While this has improved with the introduction of hybrid disk arrays utilizing flash storage it isn’t moving at the rate necessary to move the needle so far as to cause companies to reanalyze their desktop deployment strategy. Why haven’t storage costs changed dramatically? Storage vendors don’t have incentive to reduce their costs…they make more when you spend more, why would they want you to spend less?

What I’ve grown frustrated by in the last year is that VMware and Citrix are not doing more to further the deployment of Desktop Virtualization using local storage. Citrix recently has even gone so far as to deemphasize their own hypervisor XenServer in favor of Microsoft Hyper-V. Microsoft Hyper-V 2012 includes features such as deduplication and a memory-based read cache. So why not go one step further to design, create, and endorse local storage solutions? A couple years ago Citrix bought Kaviza, a grid solution built around using local storage, but where did it go? It seems that the only companies getting rich on VDI are the VDI supporting vendors…most notably the storage vendors and I don’t understand the behavior of the Desktop Virtualization vendors any longer.

It’s time to embrace local storage. It’s time to stop debating the macro features that only matter to 1% of the world. It’s time to focus on how to make Desktop Virtualization valuable to the 99%. Some things can’t be solved, others can…cost is one of those things and local storage is the key. People are lemmings and want to know that others have gone this path and that their vendor supports them before stepping out and deploying technology in a new way. It’s time for Desktop Virtualization vendors to step up and lead rather than let their market success be controlled by the 3rd parties making more on Desktop Virtualization solutions than they are.

While we’re at it…fuck you Microsoft VDA licensing, let’s see what happens when companies start deploying server OS 1-to-1 for the sole purpose of avoiding this tax. #fixVDA http://blogs.gartner.com/gunnar-berger/attacking-the-high-costs-of-desktop-virtualization-part-1/  Fix VDA and start using local storage? Maybe there will be a “year of VDI”

9 thoughts on “Moving the needle on VDI

  1. Couldn’t agree more. We, be that users/enterprise’s are continually held hostage to the vendors. Fair enough they need to make money, but many egotistically refuse to co-operate, and cynically tell us that “you need xyz” which is their way of doing something the opposite of the other vendors approach, and thus stubbornly refuse to modify their products beyond the one side of the argument(s) they have taken. I think Harry Labana said “we need co-operation not competition”, and I’m sure we would be 20 years ahead of where we are now if there was.
    There will never be a “year of VDI” until the minutia of differences get resolved, or a vendor leaves their ego at the door and provides a product with flexibility to provide what we want (that vendor would clean up by the way, and their “year” could be a decade).
    Its coming but whilst we are still buying their current “stuff” vendors wont fix these issues. Its only if we stop that they will – maybe we need a “user trade union”!

  2. Local spindles seem great but there are drawbacks like the loss of vMotion and HA. An even better solution would be to use local SSD as a caching device so you can offload storage IOPS (both reads and writes) to local spindles without compromise. Write caching to local SSDs is only recently becoming a reality. Example products include: Nutanix, PernixData FVP, VMware VSAN, and Fusion-IO ioVDI.

    1. Non-persistent desktops dont need things like vMotion and HA. With persistent desktops it’s arguable. Laptops and desktop have probably a 24 hour SLA for repair…at least..yet we move these workloads into the datacenter and automatically give them 5 nines of availability? Who is accounting for these costs?

      Spindles are being replaced by SSD for reads and writes and we’re on the verge of consumer storage entering this area. http://myvirtualcloud.net/?p=5284

      Nutanix, PernixData and others have made storage cheaper but it’s still at a significant premium to the raw costs…and rightfully so, these guys need to make a living too.

      I am suggesting that local storage VDI solutions should be based on consumer or prosumer SSD, no storage vendor involved at all.

  3. Dan,

    I wrote an article entitle “Welcome to the new Transient Desktop deployment model” where I pretty much agree with your take, but I also propose a solution with Horizon Mirage. http://myvirtualcloud.net/?p=5215

    Not trying to market the solution here, but it is a simple and effective way to provide persistency to non-persistent desktops.

    -Andre

    1. No doubt technologies like Citrix Personal vDisk and Horizon Mirage have the ability to get us to the persistent desktop with non-persistent design and underlying infrastructure. Big fan of Mirage here too 😉 SSD and more IO than we need for a virtual desktop gives us opportunities to do things we couldn’t consider before…things that were too IO intensive. Will be interesting to see how Mirage comes to market on non-persistent virtual desktops.

  4. Hi Dan,

    I would be real curious on your take on Dell vWorkspace because in a Hyper-V environment this uses local disks almost in the same fashion as VDI-in-a-Box. So rather than rely on complex or expensive systems for cloning in bulk, copy an updated master template to each node, then start the cloning within each node.

    This eliminates any need for discussion or calculation of IOPS, all the hard work is now contained within each separate node.

    Cheers,
    Dave

    1. I haven’t personally done anything with Dell vWorkspace but I do know they have done more around local disk architecture and intelligent placement of new connections given the lack of vMotion/Live migration when using local disk. I haven’t heard much else about Dell vWorkspace, they might very well have a great solution but they have a very small part of the market today when compared to VMware and Citrix.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s