Justifying the Cost of Virtual Desktops

Recently I was putting together a proposal for a customer who needed to refresh 40 desktops, in the proposal I needed to compare the costs of replacing the aging desktops with new HP Desktop PC’s versus replacing them with VMware View Desktops. This particular customer already has VMware and has the need to upgrade their SAN and cluster capacity for another project, so adding in a little more disk and a little more compute resources would not be an issue. Instead, they wanted to know what the true cost of licensing, hardware and other indirect costs were on both sides.

What I came up with was about a 30k dollar delta to go with virtual desktops instead of physical HP Desktops. Obviously, I want to push for virtual desktops because of the lower maintenance involved with maintaining the hardware, but I would still need to justify that extra $30k. The first thing that I used was power.

For 40 HP desktops, I figure they would pull about 200watts per desktop if those desktops run 24 hours a day for a month that is 733 hours. That is 146.6 kWh per desktop and I figured 10 cents per kWh as the cost for power, which makes each desktop cost about $14.60 per month to run. Multiply that by 40 desktops and we are at $586.40 to run our desktops every month.

Now let’s look at virtual desktops: I am suggesting the EVGA PD02 terminal which pulls on average 12watts (according to their documentation), I then multiplied it by 733 which gave me 8.79kWh per month and then I multiplied that by 40 terminals to get 351.6 kWh per month… for all 40 terminals!!! That is $35.16 to power them, I also added 200watts for the extra server that it would take to run the desktops which add $14.66 to the monthly cost… for a total of $49.82.

So overall going with terminals and View Desktops will save us $536.58 per month, or $6439 over 12 months… which works out to $19317 over 3 years!

What I’m getting at here is that the power savings alone will be enough to pay for 2/3rd’s of the upfront cost of VDI over 3 years. To finish justifying VDI… figure out the amount of time it takes to update one application on 40 PC’s…. then figure the amount of time it would take to update one template and refresh the desktop pool…. Over three years of application installs and upgrades It is quite clear that VDI will more than pay for itself!

Also, on a side note, I would expect that 10% of the terminals over 6 years would fail… But over 6 years I would expect to replace 75% of the PC’s…. so If you were to figure a 6 year replacement schedule VDI really shines as the true lower cost.

Also, check out my “Take 2” post on justifying VDI.

Share This Post

12 Responses to "Justifying the Cost of Virtual Desktops"

  1. I know the comparison depends a lot on which models you select, but we recently replaced ~2000 desktops with new HP DC8200 SFFs, and we found their powerdraw, under a normal workload to be around 20-30W. IIRC that’s with a Core i5 CPU and 4GB RAM. Modern computers can be very power efficient when they want to be. 🙂

  2. What about vmware licensing costs? I don’t know why Microsoft, vmware, and so other companies always forget to mention the licensing costs (and their yearly renewal).
    I still don’t like this figures made on-the-fly.

  3. Ya the examples are more or less useless, where can you find an office pc that draws 200watt??

    And why run it for 24/7, just make it shutdown via policy…

  4. It’s certainly unusual to run an office PC at full power for 24 hours a day. I’ve also never found it time consuming updating applications – a good software management tool should cover most applications. The majority of our applications are browser-based now anyway.

    I looked into virtual desktops in our office of about 100 users and these were my immediate conclusions:
    http://carnivalboy.wordpress.com/2011/07/26/desktop-virtualisation/

    I’m looking forward to reading your breakdown.

  5. there are a lot of technologies to deliver application on the clients without VDI…
    Data virtualization is a simple thrick, application virtualization is more expensive but still easy (on windows networks, of course).

    I think that today VDI is an expensive toy…

  6. There definitely still needs to be a business use case. But if you are able to save a bunch of man hours on common update tasks then it could be a really nice option.

  7. HE is right. It is a lot cheaper if you do it like that. But if he ever wants a newer copy of office or windows he will need to rebuy the licenses. Where as with open licenses (which cost more) you automatically have the right to upgrade as long as you have software assurance. So it comes down to how cutting edge you want to be.

    If you like having the newest versions of windows and office then open licenses with SA is the way to go. If you only upgrade every 7 years though then OEM or non-SA licenses are probably better.

  8. “If you only upgrade every 7 years though then OEM or non-SA licenses are probably better.”

    I work on a 4 year cycle of PC/Office replacement. Very little changes in Microsoft products over a four year period. They’re evolutionary not revolutionary.

  9. Ok,

    but replacement of pc/office doesnt mean savings for company (power, easy administration… everything what zero client brings into…)

Leave a Reply