An important decision to take when planning an ESXi deployment is the choice of the hardware platform of the server. ESXi doesn't support all the hardware available on the market (storage controllers, NICs, and so on) and has some restrictions that can prevent the successful installation of the hypervisor.

Only tested and supported hardware ensures that your ESXi can be installed without any problem and can operate as expected. Before purchasing the hardware for your server, it is strongly recommended you verify whether ESXi supports the chosen hardware platform.

To check for hardware compatibility, you can refer to the VMware Compatibility Guide available at https://www.vmware.com/resources/compatibility/. The list of tested hardware is large, and you can find the supported hardware from the leading manufacturers, such as HP, Dell, IBM, and Cisco.

When new hardware is released and certified for compatibility, the list of supported vendors is updated accordingly.

Choosing the right server for your installation is not an easy task, especially if your environment grows quickly and the business requirements change frequently. Capability, scalability, availability, and support are the elements of the server you need to evaluate carefully to be sure the final choice fits in the available budget without affecting the global design.

In some scenarios, it is better to have more, smaller servers in a cluster to provide the required resources than a few big servers.

There are several considerations you need to think about:

• Fewer servers with more resources are usually cheaper compared to the same amount of resources distributed among more smaller servers (you always need to buy a chassis, a motherboard, and power supplies for every server).
• With more smaller servers you can have multiple fault domains. For example, if you have 10 ESXi hypervisors in four racks, when one rack (fault domain) becomes inaccessible, you still have other fault domains compared to 15 powerful servers in a single rack.
• When your ESXi server becomes unavailable, you lose part of the computing power. If you have 10 large servers, you will lose 10% of the total capacity if you encounter downtime. If you have 50 smaller servers, you will lose only 2%.
• With more servers, you will need adequate network infrastructure concerning the physical 1/10/40/100 GbE ports.

The challenge is to find a server that provides the number of resources that meets the requirements but at the same time supports enough expansion (scalability) if the demand for resources grows.

Another factor you should consider is the expected performance of the server. The default hardware BIOS settings of the chosen hardware do not always ensure the best performance. To optimize performance, you should check some of the following in your server's BIOS settings:

• Hyperthreading should be enabled for processors that support it.
• Enable turbo boost if your processors support it.
• In NUMA-capable systems, disabling node interleaving (leaving NUMA enabled) will give you the best performance.
• Hardware-assisted virtualization features, such as VT-x, AMD-V, EPT, and RVI, should be enabled.
• Consider whether you should disable any devices you won't be using from the BIOS (Serial/Parallel ports, unused PCIe cards, and so on).
• For power management, you can choose to enable max performance or leave the control in ESXi with OS Controlled mode.