Predictable VDI Performance From PoC to Production

One of the challenges of a large infrastructure project is ensuring that the proof of concept (PoC) and pilot phases represent the final production rollout. For a VDI project, the challenge is to make sure that performance remains acceptable through rollout. One of the benefits of a scale-out solution like SimpliVity is that performance per desktop remains constant as the rollout progresses. You can be certain that desktop VM performance will be the same for the 10,000^th desktop as it was for the 10^th.

With a conventional infrastructure, the capacity and performance of the storage are completely separate from the compute (CPU and RAM). One storage array is shared by all of the compute, with a fixed storage performance and capacity. That fixed performance will be available to 50 pilot users and then later shared by 1,000 production users, so as the desktop count increases there is a risk that the storage array will become overloaded. Once the array is overloaded, every single desktop and every server VM that shares the array will perform poorly. The array that was great for 200 pilot users might not cope with 2,000 production users. It's not a fun time when Monday's new VDI logons don't finish until Tuesday.

A hyper-converged platform puts storage capacity and performance in every node. While the storage is shared, it is not fixed in performance or capacity. When you add a SimpliVity node to a cluster it brings compute resources as well as storage capacity and performance. A four-node cluster will deliver twice the total storage performance of a two-node cluster. An eight-node cluster will deliver twice as much again. Essentially, each node is a building block with capacity and performance for a fixed number of desktops, both in compute and in storage. If the pilot shows that two SimpliVity nodes can handle 400 of your office workers, that is 200 desktops per node. You can be sure that 10 nodes can handle 2000 of those office workers with the same performance.

One reason for a VDI pilot is to work out what resources your users need to get their job done with suitable performance. Once you arrive at a number of users per node, you will know how many nodes you need for your total user count. You can even start with a small number of nodes while your user count is small, then buy additional nodes as your rollout progresses. With a SimpliVity hyper-converged system you can be sure that performance for your desktops will be the same from PoC through pilot and all the way into full production.