Organizations considering adopting VDI should pay special attention when choosing the hardware to deploy its desktop virtualization platform, since the performance of the virtual desktop is directly linked to the performance of the underlying hardware.
First of all,you should make sure that the hardware is certified to work with the VDI software. Some of the major providers perform tests with several hardware components to ensure that they work correctly with their VDI software.
On the other hand, there is a big difference between compatibility and performance. If a piece of hardware is listed in a VDI vendor hardware compatibility list, it provides a guarantee, but it is not entirely sure that hardware will support a production workload. Therefore, you should determine what hardware resources need the virtual desktops and make them available.
It is also advisable to make sure that the hardware is fault-tolerant. For example, the fail of one of the VDI servers can cause a major disruption, unless you have implemented redundant hardware. We normally focus on establishing redundant hypervisor, but it is recommended that other components of VDI, as connection brokers or even DNS and DHCP servers, are also redundant.
There are certain times of the day when users consume more resources and, therefore, you should ensure that hypervisors have enough hardware resources to handle the demand for multiple simultaneous logins. In addition, some resources will have to be adjusted so that the hypervisor can use them. Like any other software, the hypervisor requires CPU, memory and storage I/O time. At a minimum, users should have 20% more than the resources required.
Most VDI projects do not have an unlimited budget, so you should invest in resources that will mean a greater VDI performance, such as memory, CPU, storage and network. If the budget does not allow you to acquire all the desired resources, the main objective should be high speed storage and storage connectivity. Meanwhile, it is a good idea to use the physical network adapters for the server, use a low-cost memory and a modest CPU.