VDI: Very Disappointed Indeed

VDI, you have been measured. You have been weighed. And you have been found wanting.

Would you be shocked to see 40,000 horses tied-up outside an automotive manufacturing facility in Michigan? Would you be equally shocked to know that your favorite VDI vendors don’t use their own technology? Before I went to VMWorld 2008, my excitement for VDI was at an all-time high. I anticipated getting there and immersing myself in all sorts of intense VDI-oriented tech discussions, going to technical sessions covering deployment, management, and security issues, and finally hitting the show floor to sniff out products and software to review for you. Though the show was heavily VDI slanted, it left me a little flat in that area. In fact, as you know from my other posts, my excitement waned into disappointment. Why the sudden 180, you ask? I realized that VDI just isn’t practical — and no, this isn’t another rail against VDI in favor of web-based OSs — it’s a genuine analysis of the technology. I think we all know where I stand on that issue but if not, please see The Short Life Expectancy of the Virtualized Desktop.

What I found from speaking to vendors in this space is that,

  1. The RDP and ICA protocols are too slow for large-scale deployment. Introducing more efficient protocols, such as Cisco’s WAAS and Qumranet’s SPICE, give hope to those whose requirements exceed that of RDP and ICA.
  2. Management applications for large-scale VDI don’t exist. Several companies, including VMware, know that VM management is a major problem and they’re working on it.
  3. Vendors admit that, even in best-case scenarios (optimized protocols, Gigabit Ethernet, and high-end (64GB to 128GB RAM, multi-core) servers), that the ratio of simultaneously running virtual desktops per single server is about 50:1. A full desktop experience (sound, graphics, video) for 50 people would still not be possible using our best-case scenario.
  4. Licensing fees remain the same, or are higher, for your desktop OS and applications in a VDI environment and licensing is a tangled web. Licensing Microsoft and virtualization products becomes very complex — are your users using full desktop OSs, Thin Client hardware, Vista or XP? Do you know what a VECD (Vista Enterprise Centralized Desktop) License is?
  5. VDI is a new way of operating the Enterprise due to the workload shift from desktop to server and that shift incurs costs. It’s a large task to manage 400 servers but once you add 5,000 or more desktops to that mix, you’ll need to add in the cost of support for the additional hardware and virtual machines, which in turn may save you very little. Remember that, even if you move to Thin Client hardware, someone still has to deploy that hardware to the user.

I’m not in the habit of beating up on any particular technology or solution and I don’t want to give you the impression that I do. Everyone believes the hype about VDI right now and it’s my job to paint a realistic picture of it and what conversion means to you and your company. You also need to be realistic about how much money you’ll have to spend to provide the type of performance that end users currently enjoy with local desktop machines — sluggish desktop performance decreases both productivity and morale.

If you still believe that converting to VDI will save you a stack of cash or that I’m just totally not in touch with this technology, then please do the following and report your results to me:

Call your favorite VDI vendor and ask them to see their virtualized desktops in action. No, you don’t want to see an online demo, a laboratory where everything works because there’s no actual load on the systems or desktops, and you don’t want to see some 50 desktop solution they’ve provided to a client — what you want to see is how they’ve implemented VDI for themselves in their Enterprise. You want to come to their main offices and see VDI in production.

You’d expect that the vendors themselves use VDI for their very own users, right? — just like you’d expect automotive workers to have cars instead of horses because they build those cars and cars are so much better than horses. I think you’ll find the answers you receive will surprise you.

Kenneth Hess is a Linux evangelist and freelance technical writer on a variety of open source topics including Linux, SQL, databases, and web services. Ken can be reached via his website at http://www.kenhess.com. Practical Virtualization Solutions by Kenneth Hess and Amy Newman is available now.

Comments on "VDI: Very Disappointed Indeed"


Well done! Kenneth’s article phrases, “Eat your own dog food!” in much more genial language.


Very interesting article. I think the key with VDI is that it can be very usefull in the right situation such as remote working or in an R&D type environment where users need a short term sandbox to play in. But as yet there is no substitute for local desktops when it comes to flexibility and performance, it will be interesting to see where things go from here.


G’day Ken. I heard Sun rolled out their own VDI internally recently. I’m sure a call to one of the local offices could confirm or refute that.


I’m very much a “right tool for the job” person. I’m mostly all about Unix in the data center and Windows on the desktop, with lots and lots of virtualization in the mix. Thus far, I’ve found only two places where VDI is going to be useful to us:
1) targeted application environments
2) multi-os clusters

As an example of one above, we have 4 people who need to use an application that is *only* supported with an insanely configured VPN. We can either deploy two machines to the offices of the users in question, or else have a VM set up for them to use for the problematic application.

For number two, think about teaching clusters. Some instructors want Windows, others want Linux, but *no one* wants to wait for the system to update itself when it boots. VDI is just about perfect for this as the VMs are completely generic and, since they’re running all the time, we can keep them updated.


A healthy dose of realism, but let’s see what happens when someone takes a crack at protocol bottleneck part of the problem. For those who are determined to go down the path of virtualized Windoze desktops, Vector Networks have ensured their stuff provides asset management functionality such as inventory and application usage measurement consistently across mixed physical and virtual environments. Software license optimization does not miraculously happen through VDI.

I suspect the same core issues of performance and resource will also prevent SaaS via the internet working for the corporate market, but that’s another story….


If only you had said what “VDI” is.


Well, my mom taught me what “VD” is, and I guess “I” stands for “inoculation”? ;-)


Sorry, Robert, VDI is Virtual Desktop Infrastructure. It refers to creating virtual machines either from scratch or from a physical to virtual conversion (P2V) from your computer to a server system. You then run a thin client, a program that connects you to your virtual machine on a server, and you run your desktop over the network.

That’s the 10 second version of VDI.

Read more about VDI on Wikipedia.

Sorry I didn’t explain it but my column/blog is sort of like a Technical Soap Opera; You just jump in somewhere and pick up the story as you go. ;-)


To Marc,

Do people still say “VD”? We said that way back when we watched those black and white school films. I think the 21st Century term is STD.


I believe Sun rolled out their own VDI (Sunray) quite a few years ago. It may be “ancient” technology and not “hip” or trendy per se, but it works..


Hi Ken,

Don’t go completely sour on VDi. From the perspective of where my current organization is with already in-place technology, VDi makes a lot of sense but it’s not a one size fits all solution.

I currently work for an organization which is a call center. Approximately 18 mos ago we went away from the ancient backend infrastructure (HP-UX riding on HP 9000 N-Class servers) with old clunky dumb terminals at the end users desk. Now we’re all terminal server on the backend and 90% thin client at the desktop. With the exception of regular staff users, we are 100% THIN at our agent desks.

Many of the challenges which require an organization to provide a Terminal Server infrastructure to its end users are similar to those of VDi. Especially when you’re talking about bandwidth requiresments and capabilties in supporting RDP over the wire. Because of this, I believe that VDi in an environment such as this can really thrive as well as minimize startup costs.

We’ve recently purchased a small amount of VDi licenses along with VDM due to other needs on behalf of our agents which are not supported in a terminal server environment. We are a moderately sized VMserver shop already with approx 60 vm guest servers in a 5 server cluster running DRS, Vmotion, HA etc…This also helped minimize costs associated with an entry level deployment of VDi.

Understanding that each organization is different, the individual items which would make VDi a success or failure will also be different. We drive base images of XP desktops to remain consistent and unchangeable. This is PERFECT for a VDi environment. With that, SAN arrays such as NetApp and EMC do some amazing things which allow us to reprovision 30+ desktops nightly throughout the organization with some fairly basic scripts and all in the matter of minutes.

In closing, I do see you point. However, I think people need to be creative in the thought process of how VDi can be leveraged today and how it can fit into environments which are difficult to support due to the physical restraints PC’s at desks.


Leave a Reply

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


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>