In a bit of a departure from my normal Group Policy banter, I wanted to talk a little bit about VDI, or Virtual Desktop Infrastructure. Like most things around virtualization, there’s a ton of hype about the promise of virtualizing desktops. Companies are starting to deploy VDI in real production environments. Its by no means the majority of new desktop deployments, but as with any new technology, there are early adopters that are dipping a toe in to discover how it may (or may not) help.
That said, my friends Don Jones and Greg Shields recently posted an article on TechTarget entitled, “Four reasons why VDI might not be right for you” in which they state right up front, “We’ll just say it: Virtual desktop infrastructure is a bad idea.”
Well, that statement is eye-catching but in my experience it is not the whole story. So I thought it would be a useful exercise to review their 4 points and see where there is agreement and disagreement.
1. Their first point is that “You might not save any time or money” and go on to say that , “ If you end up having one virtual desktop per user, then you have achieved nothing other than relocating those users’ computers from their desks into the data center”
My own experience is generally different here. First off, deploying VDI is not always about saving money, as strange as that seems. Sometimes it is about security–moving certain user population’s desktops to the Data Center solves some otherwise thorny data loss prevention problems that are critical to many organizations. Throw in the fact that software distribution and patching stays within the data center and is far more likely to have higher success rates, and its easier to make the case based on that centralization alone. The comment about ending up with one virtual desktop for each user is important, but misses the larger point about VDI. Let’s look at where desktop management costs come from first. Primarily you have your capital costs (hardware and software) and then you have your operational costs (people costs to maintain systems). Most would agree that the operational costs of the typical desktop far eclipses the capital costs, but those operational costs are often hard to quantify, so people often fall back to comparing capital costs. Ok. So is VDI always more capital intensive than physical desktops? Actually, not always, even in the scenario where you have 1 virtual desktop for every user, depending upon the workload, density of virtualization on your hypervisor, etc. I have seen examples where the costs are roughly the same between VDI and physical hardware. However, I do want to acknowledge a key point that Don and Greg make here. If the “access device” you are using to access your VDI instance is just another PC running Windows, then it becomes harder to make the case that patching and updating your access device and patching the VM is ever going to be cheaper. This makes a stronger case for considering thin terminals or technology like WinFLP (it has a new name now that eludes me at the moment) as the preferred access device for VDI.
2. They state that, “Virtual desktops can cost more than real ones“.
This is a continuation of #1 above and indeed there is the risk of this, depending upon how VDI is deployed. I think the first incorrect assumption here is that most enterprises are paying 400-500 for desktop hardware. While you and I might pay that for a stripped down PC for the home, few enterprises buy such machines. More realistically, enterprise PCs can typically cost anywhere from 600-1000 dollars or more, depending upon the bells and whistles. At the higher end of that range, VDI gets very close, especially when using blade servers with integrated storage and networking and again, depending upon the densities you achieve (i.e. # of VMs per physical server).
They go on to talk about the so-called “false costs” of managing desktops, “You can ignore them in the calculation above because they don’t go away with the move to VDI”. However that is only the case if you deploy one persistent desktop to each user, which is arguably the “training wheels of VDI”. They hint at this by talking about pooling, but pooling is only half the story. The real benefits of VDI come in when you deploy technology that lets you share a base image across multiple users (e.g. Linked Clones from VMWare or the OS Streaming technology in Citrix XenDesktop). This is where the costs really start to make sense because not only do you start saving big time on storage (admittedly the largest chunk of VDI HW costs) but you also can potentially reduce your management costs (those “false costs”) by only having to patch one desktop image for every 100 or more users.
3. This point talks about storage, where they state that, “VDI converts inexpensive desktop storage to expensive SAN storage“
I would agree with this if SAN Storage were the only option for VDI and if someone wasn’t deploying one of the image sharing options I mentioned in #2. But again, that would not be the best practice. When you also take into account the fact that user data is already typically redirected to file servers, that applications can and should be virtualized and that you no longer have thousands of DAS volumes to try and backup, protect, and prevent information leaks from, the hard and soft costs of storage start to work in VDI’s favor.
4. They wrap up by saying, “VDI’s layers are still too many, too complex”
While I wasn’t able to find the presentation they referred to that detailed 19 different layers to VDI, my experience has not shown that. Aside from the connection broker and client access pieces, many of processes enterprises use today to manage servers and desktops is applicable to VDI. I think one key challenge in VDI is mostly organizational. That is, many enterprise shops have traditionally separated desktop and server into separate groups and disciplines. VDI forces those to come together and sometimes that can create challenges. I would wager that for many shops, overcoming those are as big as the technical challenges.
Overall, despite the fact that I have some disagreements with Don and Greg, I do agree that their overall premise, that VDI is not a slam dunk for many shops, is spot on. It certainly doesn’t make sense yet for most IT shops that aren’t enterprise in size. Even enterprises today are walking before they run, with VDI, and are probably not delivering on the promise of VDI’s potential cost savings. In addition, not every user workload should be considered a VDI candidate. There are some workloads where the costs just don’t make sense, where technologies like Remote Desktop Services or, good old physical PCs make more sense. But longer term, as companies and technologies evolve, I believe that VDI will deliver on its promise, just as server virtualization has, and will be a dominant presentation mechanism over time. Throw in the fact that VDI frees you from having to use a PC as an access device (think iPad or other mobile device) and the possibilities grow exponentially.