We have this great cost savings (TCO) story on desktop virtualization. I’ve got the PPT, the PDFs, analyst quotes, the whole package. But I am finding that before I ever get to talking to customers about cost savings, some genius in the room brings up the obvious question: What’s the point of getting the TCO pitch if there is no budget for virtualization servers, thin clients, and Microsoft licensing? It’s a good point: this stuff is pretty expensive to set up in the data center. My answer is to look no further than their existing PC replacement budget. I bet all of you have a corner of the office building that looks like this photo below. People are throwing out these things on a 3 or 4 year cycle which, for you math-challenged, is 25%-33% per year. They have to be replaced by something; generally it’s with another PC that looks pretty much the same as the old boat anchor to the tune of $600-800 each!
For me that’s money that should be spent on desktop virtualization. For most vendors, desktop virtualization solutions end up costing more than $800 a person when you add up the cost of licensing, thin clients, virtualization servers and Data Center upgrades. So, even if you spent the entire desktop replacement budget, you will only be updating 20-25% of the end users, leaving the majority of users missing out on TCO benefits. Kind of a bad deal when you think about it.
Some of the bright guys at Citrix came up with an idea to centralize ALL of the desktops while staying within the current year’s PC replacement budget.
First, you delay replacement of the endpoint by using the processing power of the endpoints to reduce the number of hypervisor servers required in the datacenter. XenDesktop streams the OS from the data center directly to more current PCs that were not slated for replacement. The “walking wounded” PC’s get repurposed to receive hosted desktops and the ones that are really, truly hosed get replaced by a few thin clients. Here’s the 5-step program guaranteed to free up those skids full of old PCs:
1. Examine existing PC inventory to determine capabilities to support streamed or hosted desktops.
2. Relatively new PCs (less than 3 years old) can run centralized, streamed desktops. Users will experience similar performance to existing environment but will benefit from faster boot times and better overall reliability. In most enterprises, this will cover up to 80% of the existing endpoints.
3. Older PCs (more than 3 years old) will be repurposed to support hosted desktops. Users will experience much better performance than before since the endpoint will only be used for display, keyboard, mouse, and network connection. The actual desktop will run on hypervisor servers in the datacenter. In most enterprises, this will cover up to 20% of the existing endpoints.
4. Non-functional or obsolete PCs will be replaced by thin clients running hosted desktops. In most situations, this would represent less than 5% of the existing endpoints.
5. As the PC’s receiving streamed desktops age, they will be migrated to hosted desktops, effectively doubling their lifespan. As these PC’s fail or become obsolete, they will be replaced by thin clients displaying hosted desktops.
Using this methodology, PC’s are slowly replaced over a 5-8 year timeframe and hypervisor servers are gradually added into the data center to support thin clients. This approach has the benefits of minimizing and spreading out acquisition costs, immediately reducing TCO, and gradually reducing power and cooling costs by introducing thin client endpoints over time.
Here’s a quick budget spend bar chart for 1000 desktops ($800 replacement cost), assuming existing 3-year replacement cycle, full use of PC repurposing along with a generous licensing discount on XenDesktop from your friends at Citrix.
This works for me. I’ve got the entire PC population centralized and I haven’t spent the entire IT budget to get there. In fact, I got there using one year’s desktop refresh budget item. More importantly, my TCO arguments apply to 3-4X as many user endpoints than they did before.
The lesson is: don’t wimp out with a small pilot; centralize the whole PC population with a quarter of the upfront cost that VMware would quote.
So, folks, what do you think of this idea of moving the entire desktop population to a centralized model? It’s sure to take a bit of rethinking around support strategy. On my next blog, I will take the next step and spend some time describing the cost savings story behind XenDesktop.