The main issue with VDI? Windows. 8


Yes, you read it right.

As the VDI debate continues, now heated up thanks to the iPad (piece of crap IMHO, subject to another post), I decided to write this post that has been sitting here, waiting for me for at least 6 weeks. It goes to the heart of VDI: Windows.

As of today when we talk about a hosted desktop solution, we like it or not, Windows is the OS of choice (the desktop versions is what we are discussing here like XP, Vista, 7). And the reason why I think VDI has a long, really long way to go, unless Microsoft takes action, is this same OS indeed. Windows.

Let me start by saying this. There are several posts and information on the web that clearly show that Windows was optimized over the years to run on, guess what, real, physical hardware. Why? By the simple fact until people started talking about VDI (circa 200X), all Windows deployments were 100% done on physical hardware! That is why the OS was tweaked/optimized to run on real hardware. Kind of makes sense huh?

Now if you look at a post by Ruben on storage  this is clearly shown and stated. And we are just talking about the disk subsystem here. There are for sure several other things/components that were changed/tweaked to get the best performance out of real hardware.

Add to that a very simple thing: Windows was never designed with things like application layering (explained on this post by Gabe), sharing a master image with differential vDisks and so on, in mind. These changes, required to make VDI an affordable, scalable and stable reality, introduce several issues. The main one for any serious, large deployment, will be what? Support. The next one, the simple fact that no single vendor offers all that is needed for a scalable, stable VDI solution. This means you will probably end up with VMWare on your virtual backend, Citrix XenDesktop as your VDI solution and several other pieces from several other vendors like Atlantis, MokaFive, McDonalds, you name it. Yes, McDonalds is jumping into the VDI bandwagon (what leads us to my post about VDI and Patchworking  – worth reading – bringing several issues to the table).

Back to the topic, even though some vendors may say their mechanisms are not that intrusive (like my discussion with John Whalen from Mokafive last night on Twitter), the bottom line is not 100% of your apps may work and more than that, if they apparently work and you find issues down the road and call Microsoft or any other vendor, chances are they will simply tell you to go ____ yourself. You can fill in the blanks.

Some may say that was the case with Terminal Services/Citrix years ago. Yes, in a way that is true. The difference is TS/Citrix was in several ways, way, WAY less ‘destructive’/’intrusive’ on its approach to make things work, than VDI is. VDI has to deal with sharing disk images, dealing with deltas for each user. Dealing with layers. And so on. If you know the internals of any OS you can see right there what sets VDI and traditional SBC apart.

As soon as Microsoft brought TS under its umbrella, making it an OS (NT4 TSE) or a service on Windows Server OSs (since it introduced Windows 2000 Server), things changed. All the sudden Microsoft had to support its own solution and products running on it. Vendors could no more ignore the fact people were actually using TS/Citrix to run their apps. And if you look at Windows Server 2008 R2 you can see how Microsoft changed the OS to make RDS (formerly known as TS) a better solution for hosting applications. Not to mention the release of tools like the RDS Application Compatibility Analyzer.

So at the end, Microsoft changed Windows Server to make it the ideal SBC platform (I will not go into discussing if they succeeded or not – I do think they have done, with Citrix, an excellent job over the years; still room for improvement, like anything else in life).

VDI is no different. The problem is we now have much deeper issues related to the OS than before. And the only ones that can actually fix these is Microsoft. Period.

Windows may need a big redesign to accomodate VDI needs/requirements. I am sure there are several things that could be changed on Windows to make it the perfect OS for VDI (what I do think most of you will agree with me, Windows is NOT perfect for VDI; for God’s sake, even on physical hardware it has its own issues). Once these changes are done (some may be fundamental changes on the OS) I am sure we will be able to scale a VDI solution without all the storage hassles, disk image sharing, disk deduplication and so on. And it will be supported and (knock on wood), stable.

Of course I assume you want VDI to be scalable, stable and supported. If you do not need all three, for sure you can deploy a VDI solution today. It will be scalable and stable but unsupported. Or scalable and supported but unstable. Pick your two options.

If you think I am nuts, go ahead and leave your comment or email me directly. The bottom line, at least for me, is simple: Windows was never designed with VDI in mind AND VDI has deeper ties than TS/Traditional SBC had at lower level OS components and these two little things introduce several issues.

And that is why, as mentioned several times, thanks to these issues that companies like MokaFive and Atlantis exist and the reason why VDI as a solution keeps moving forward (honestly, I think we all owe a lot to these guys, the ones trying hard to make the virtual world less virtual, more real).

This is simply put, people taking matters into their own hands, while we wait for the day Microsoft will release Windows-V, the first release tailored for virtualization.

Windows-V? You heard about it here first.

CR

10,498 total views, 1 views today


Leave a comment

Your email address will not be published. Required fields are marked *

8 thoughts on “The main issue with VDI? Windows.

  • appdetective

    I disagree with your assertion that you need all the sharing crap to do VDI. You need that if you are short term TCO focused which most people are. These are the same people that don’t get why people do VDI, centralization etc. It’s more than XA also, but few understand why VDI vs. XA etc.

    What’s needed is a next generation of management thinking to enable the use cases. MS will never do it, because they have no incentive. MS is exactly the company that will do nothing, they will only buy into the game. Hence we need the startups and the Citrix’s and VMWares of the world to innovate to force the MS hand.

    • crod

      I am not saying you need all that to do VDI. The same way you can have a TS environment with a single TS, you can do plain VDI with nothing else. 🙂
      The point is if you, as of today, do not use several other things, VDI may become a very hard proposition. For example, with no shared storage/deduplication, an environment for 35,000 users may need 350TB of disk space. There are several other examples that show you what happens if you try ‘plain’ VDI and what the requirements become (not to mention management tasks/issues).
      Also I am not saying Microsoft will change whatever is needed. Nope. But they are indeed the ones that hold the key for VDI’s future. The more problematic/complex it becomes to deploy a hosted Windows desktop, the less mainstream it will become.
      I do agree with you it all gets down to needs/requirements/business case. This does not eliminate the technical flaws highlighted on my post though.
      Thanks as always AppDetective. Really appreciated.

      CR

  • appdetective

    Even with 350TB of central storage 1-1, it’s not the end of the world due to the benefits of centralization etc. However I do agree that without the other stuff you refer to, it’s a niche use case and not mass adoption (that is VDI). I don’t believe MS will ever solve it, so we’ll have to see how far this thing goes and I think many of the problems will be chipped away at by other because of the size of the pie. It’s also very possible that once people get over VDI hype and PCoIP horseshit marketing, they will understand the limitations and simply conclude XA on RDS is good enough to address all the structured task worker use cases, and cheaper. The rest will be fat PC’s and I think here outside of BYOPC client side hypervisors have a role to play to reduce mgmt costs period. Again I expect MS to do fuck all here and it will be up to the other.

    • crod

      That is it. Now we are indeed on the same page. 🙂
      To be mass adopted and leave the ‘NicheWorld’ VDI needs fundamental changes on the OS and maybe at the hardware level as well (my post about ‘Intel buys Citrix’ is a good example of that). The other thing to be considered is how a virtual world may at the end reduce the sales of what we know as a ‘PC’ and how that affects the revenue stream for Microsoft (meaning if it is all virtual and I can use an iPad like device to get to my ‘virtual desktop’, why am I going to bother to have my own PC running a local OS? I can access all my different VDI instances from a single device, screwing up PC sales potentially). The battle is not only technical. It is political and probably with VERY big fishes in the pond…

  • Daniel Feller

    Interesting read. I don’t think VDI-only is the way to go. I’ve stated this many times. I don’t think XA is the only way to go either. There are niches for every user and use case. Some of us require the mobile device, while others need a completely locked down device because they are the task worker/data entry person who just does things over and over again. The question you have to figure out is if a desktop virtualization solution is going to buy you anything? Will it be easier to manager, update, patch, support, etc? If your analysis is no, then time to move on and figure out another way to fix the nightmare that is the desktop environment.
    Twitter (@djfeller)

    • crod

      I do agree Daniel and more than that, that is exactly what I have been saying for years. The reality is several analysts and even Citrix in certain ways sees VDI (or is positioning it) as a physical desktop replacement alternative. Just look at how XenDesktop 4 was licensed initially and also how Citrix, marketing wise, is trying to separate XenDesktop from the ‘old-school’ XenApp (SBC) way of doing things. If we think market wise it does make sense as the desktop market is way bigger than the SBC one so if you position your product as a solution for desktop needs for sure if you get 1% of such market is probably a 100 fold increase compared to traditional SBC.
      Resuming: technically we both agree VDI is no silver bullet and must be used, like any other technology, to address particular needs. But as I said, several people in the industry are positioning as the ‘God sent’ solution to save us all.

  • appdetective

    And it is because of this disruption so many possibilities will open up and why I am a huge fan or all those playing in this game, despite give them hell for not being fast enough to innovate or feature complete. Key will be quality execution.

  • Harry Labana

    Crod, the problem is that people misinterpret what Citrix is saying. We are not saying VDI is a silver bullet and the SBC way of doing things is perfectly fine as part of a Desktop Virtualization solution. In other words options to address different use cases. We call that Flexcast options. Also to keep this balanced, there are plenty of customers who are using the VDI model as a full desktop replacement, because they want the single OS session not to themselves. Nothing wrong with that either. Too much confusion out there, that a single sledge hammer fixes all. I think I need to blog about this. It’s your fault if I write one in the next day or two 🙂