As a rule of thumb, never overcommit memory in a VDI environment. This can have many negative knock-on effects if memory is not granted when required, which will ultimately affect the end user experience.

When it comes to CPUs, while it would be nice to also not to have an overcommitment, this would simply not be affordable. CPU overcommitment, if done carefully and not pushed too far, can usually be allowed with little to no effect on the end users. However, the question is, how far is not too far? This will generally depend on the type of workload you are running within your environment. If you look at various resources on the internet, you will find different answers to this question, with some claiming figures of more than 10 virtual CPUs (vCPUs) per physical core. The only true way to find out what is going to be acceptable in your environment is by reviewing the CPU Ready figure; you can review this metric via vCenter, ESXTOP, or similar tools. When reviewing the CPU Ready figures, you should initially be looking to ensure that you are keeping CPU Ready below 5% per vCPU for the desktops in your environment. Your environment might be able to accept CPU Ready higher than 5%, but this should only be after testing during your POC and pilot stages. Generally, if CPU Ready is as high as 10% per vCPU, the environment is going to struggle enough so that it affects user experience considerably.