We weren't supposed to use Windows 7 x64 with VDI. If that's the case, why are so many doing it?

I was surprised to see that over 34% of people are using Windows 7 x64 as the guest OS in VDI environments!

Last week, Project VRC released the results of their State of the Industry Survey, and upon reading it I discovered several interesting things (which I wrote about last week). Besides the fact that significantly more people are doing non-persistent VDI than just about everybody thought, and that only about 50% of the organizations are using something beyond scripts and GPOs to manage their user environment, I was surprised to see that over 34% of people are using Windows 7 x64 as the guest OS in VDI environments! This revelation goes against just about every best practice I can recall hearing, and I can't imagine 34% of the organizations have uses cases that demand 64-bit OSes. I want to look into the implications of this a bit, in hopes of either discovering why companies are doing this or educating as to why they shouldn't.

Click to enlarge

From the survey, we can see that themajority of people are using Windows 7 x64 on their systems, followed by 29% of people running Windows 7 x86 and 25% (gasp!) on Windows XP x86. The Windows XP number surprises me because you'd think that if you were moving to VDI, you'd be upgrading your OS as well, especially with the end of Windows XP approaching in just over a year. That is an article for another day, perhaps, but suffice it to say that those 25% should be getting their migration in gear quickly, either to Windows 7 or to Windows Server 2003 R2.

Windows 7 x64 is in widespread use on physical desktops in organizations. The vast majority of people that I talk to are replacing Windows XP desktops with new hardware running Windows 7 x64. In fact, it seems most of the people that aren't deploying the 64-bit version are doing so for some application compatibility problem that requires 32-bit (for instance, the need to support 16-bit apps or installers). Of course, there are other solutions for this, but at least they got to Windows 7. On the desktop side, hardware is cheap and the enhancements are valuable, so the impact isn't noticed or cared about.

What is that impact, though, and how does it translate to VDI? It's commonly tossed about that Windows 7 x64 consumes somewhere around 200MB more memory than it's x86 equivalent, and if you multiply that by your VDI users, this can add up quickly. If you have too little memory in your machines, you're paging more often and require more storage resources, and if you address it by adding memory, there's a cost associated with that, too. VMware's own Server Storage Sizing Guide suggests that Windows 7 x64 VMs require 2GB of memory (as opposed to 1GB for x86), 2 vCPUs (as opposed to one for x86), and an additional 4GB of storage space. Those numbers effectively cut capacity in half, while increasing storage requirements.

What we don't know, based on the survey, is the amount of memory these machines are using, and if they've been deemed to require more than the 3GB of memory that a 32-bit machine can address. If that's the case, then 64-bit it is. We also don't know the breakdown of use cases, applications, or how many Windows 7 x64 VMs are used. The question was simply "Which OS is/are used for the desktop VMs?" So it could be that 34% of organizations have just a few x64 guest OSes, but if that were the case I'd expect to see far more than a 29% share for Windows 7 x86. We also don't know if the organizations are using mostly 32-bit or 64-bit apps, which would be important since there is more overhead using a 32-bit app on a 64-bit machine.

Based on the survey, it appears that most organizations do at least some sort of testing ahead of time, so I would imagine that most of the implementations of x64 Windows 7 have done so knowing the impact it has on their systems. Perhaps the real-world resource implications aren't as large or impactful as is commonly believed. It could also be that the experience or performance benefits outweigh any costs associated with added capacity. Of course, it could also be the "It's what we do on our physical desktops, so it's what we do in our VDI desktops" mindset, too. In a way, that might make some sense. Rather than having two platforms to worry about, you'd just have one, and aren't we at least somewhat willing to pay for that convenience?

This year at both BriForum London and Chicago, Ruben Spruijt and Jeroen van de Kamp are giving a best practices and performance session using results obtained by Project VRC, so hopefully that will shed some light on the situation. I've yet to see a comprehensive, multiple platform comparison between the two, so I'm looking forward to that session. In the meantime, what do you have to say? Do you use x64, or x86? What has your experience been with either?

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

@Gabe I'll share a dirty little secret. People don't implement VDI for cost savings. :-) Don't know why we keep going down this rat hole year after year. Sure it's getting cheaper with technology but that's not what people who are successful at scale lead with.

x64 in many cases is an incentive to get people to move to a more powerful desktop. I can certainly attest to the fact that I told my users that this new more powerful desktop for their resource hungry apps was only available this way. I have spoken to many people who have been highly motivated to get to Windows 7 and Window 2008 R2 to improve app compat between XenApp and VDI XenDesktop implementations. This is because the Windows Kernel versions were the same after years of mismatch between Windows server and client. As this was thought through, since Windows 2008 R2 is x64 only, it made sense for many to further reduce app compat testing by just testing for a single x64 bit architecture and treat x32 etc. as exceptions.

For many, it makes sense to push the x64 offering on Windows 7 to reduce overall operational complexity, and at the same time offer a benefit to the end user/in-house developers. Carrot and stick as they say... and worth it for a little bit more memory that get's cheaper over time and the benefits of having a desktop architecture that can scale to future app needs. I.E Cost is not the driver, it's all about capabilities and investing in them in an efficient way.

Also for the record I don't think the survey is any indication of persistent vs. non persistent. There is just no way people are succeeding with this at scale. Happy to argue about that one at length. The question for each of these so called successes is to understand the percentage penetration of total user base and then to understand the variance of user types. Only then can you get to truth and appreciate the reality of how different true personal desktops are vs. niche use cases with today's capabilities.  


@Harry - The issue is the survey doesn't ask about the complexity of the image deployed via non-P VDI.  So if a customer has 1000 seats of call center workers running 5 apps, it qualifies as lots of non-P VDI.  But what percentage of total desktop footprint is that 1k users?  That's a better question to ask than what % of your VDI is non-P vs P.



@Shawn, We're saying the same thing :-) I argue about scale because as we all know that's where user diversity increases and things become more complex. It's easier at that point, to use local storage or other tech to reduce storage footprint and as you say in one of your other comments today, stick to a PCLM tool to manage across your physical and virtual environments and then go have a drink!


@Harry - I know we're saying the same thing.  I was merely adding that the high amount of non-P is easily explainable if you focus on the types of VDI where it's being used.  When you silo everything it's very easy to throw it to RDS/XenApp or non-P VDI.  When you try to make that your standard model that everyone falls under, it's begins to show it's cracks and fall apart.  That's why I've been pushing persistent VDI since the beginnings of VDI despite everyone telling me I was nuts and it doesn't make any sense.



So from an enterprise perspective the VDI os is never selected in isolation. Vast majority of enterprises will have XenApp and will be running Windows 2008 R2 which is obviously 64 bit plus new physical machines are 64 bit, so for conformity then selecting win 7 64 bit makes total sense and in theory lower support costs for both the OS but also application compat, plus it's the future why delay the inevitable?


Here is my opinion.

The problem is that nowadays applications uses a heavy amount of RAM.

Even for office use you have good chance to be short with 3GB of RAM.

Moreover managing 32 and 64bit images and applications is a non-negligible task to take in consideration.

Here the question is : does 200MB x Nbr of user really matters vs time consummed building and maintening several images and applications?

There are compelling logical reasons to use x64 for VDI.
1) It's trivial to add more memory above the 4GB physical limit that x86 Windows can't address, should workloads require it.
2) Managing and testing both x86 and x64 platforms is expensive
3) 32-bit apps get more memory on x64, so the user experience is better.