How should IT approach VDI provisioning?

How to approach virtual desktop resource provisioning

A good first step for getting resource allocation right with virtual desktop provisioning is to look for manufacturer recommendations. Assuming users work with a well-known desktop operating system such as Windows, the virtual desktop vendor IT uses probably provides guidelines for hosting that OS. The OS vendor will also likely provide guidelines pertaining to the OS's hardware needs.

Keep in mind, however, that some applications are very demanding and require hardware that exceeds the minimum requirements of the OS. As a result, it's a good idea for IT pros to check the requirements for any especially demanding applications they plan to run directly on the virtual desktops. Remote applications, however, don't run directly on the virtual desktops, and therefore do not matter as much.

Another thing IT pros can do when VDI provisioning is to create a lab deployment with a VDI host and some virtual desktops. This gives them the opportunity to experiment with hardware allocations without the risk of affecting end-user sessions.

The host needs some memory and CPU resources of its own to perform VDI-related functions.

If IT pros want to avoid both under and overprovisioning resources, then they won't be able to rely on subjective estimates as to how well the virtual desktops are running. Instead, they must use performance monitoring software to quantify the operating system's performance. This is the only reliable way -- short of purchasing a third-party tool -- that VDI provisioning can determine exactly how the OS is performing and how it is using the hardware resources IT made available to it.

As IT pros dial in on the appropriate hardware allocation during the VDI provisioning process, they should leave some of the host's resources unallocated. The host needs some memory and CPU resources of its own to perform VDI-related functions.

Likewise, the virtual desktops should run at a level that does not completely deplete the underlying storage array of IOPS. The extra IOPS are necessary to maintain performance whenever IOPS storms occur.