As if paying a lot more per GB for your storage isn't bad enough, you are also going to need more of it. If your administrators use their experience configuring storage on physical servers to estimate their requirements in VDI environments, you will generally be surprised to find that you need to purchase 30-50 percent more storage to meet your performance requirements.

Why does this happen? Basically, the I/O patterns generated in VDI environments, where you may have 50-70 virtual desktops, each with their own individual I/O streams, running on a single physical server, end up being significantly more random and significantly more write-intensive than they are in physical environments as you write them down through the hypervisor and out to disk. Spinning disk performs at its worst in very random, very write- intensive environments, with the slowdown being worse the more random and write-intensive they are. For a given performance requirement (e.g. I/O's per second or IOPS), the storage configuration that met your needs on a dedicated application server will appear to run at least 50 percent slower in a VDI environment, and often much more. As administrators add more hardware (e.g. more disk spindles, solid state disk, etc.) to get back to their original performance target, the costs mount. You may meet your performance goals before you run out of storage budget, or you may not, but either way, you end up spending a lot more on storage than you probably originally planned.

If you're like most IT shops, at least part of the reason you're looking at VDI is to decrease the management and administrative costs associated with tasks like patch management, upgrades, and enforcing some level of standardization in the desktop tool sets. Moving to a centrally-managed environment where virtual desktops are served out on demand can make a huge impact here, but storage administration is almost always an area where costs increase.