The number of hosts that are required for your Horizon View infrastructure is usually dictated by the number of desktops required, the amount of CPU and RAM these desktops require, what overcommit ratio you can allow for the CPU within your infrastructure, and how much CPU and memory you can physically fit into your chosen host servers.

Taking that into consideration, you should be looking to include the amount of memory and CPU cores across the infrastructure that allows you to balance these in a cost-effective way without too much wastage.

When selecting your host server platform, you should also consider what effect it would have on the business if that one host was to fail. As such, sometimes, you might consider hosts with two physical CPUs to be a better design decision than having four physical CPUs, especially as the number of cores per CPU continues to increase.

This may well become a financial consideration as well as a technical one, and introduces the scale-up or scale-out argument of should you have fewer, larger servers, or spread the load across more, lower spec servers.

Within your calculations, ensure that you are considering the overheads that the ESXi hypervisor requires to be able to run your virtual machines, as well as memory to be dedicated as graphics memory to virtual machines if required.

The following table details some typical overhead values (in MB) that are required per VM:

Next, we will look at the networking considerations.