Plan the configuration of the hosts in the vSAN cluster for best performance and availability.

Memory and CPU

Size the memory and the CPU of the hosts in the vSAN cluster based on the following considerations.

Table 1. Sizing Memory and CPU of vSAN Hosts
Compute Resource Considerations
Memory
  • Memory per virtual machine
  • Memory per host, based on the expected number of virtual machines
  • At least 32-GB memory for fully operational vSAN with 5 disk groups per host and 7 capacity devices per disk group

Hosts that have 512-GB memory or less can boot from a USB, SD, or SATADOM device. If the memory of the host is greater than 512 GB, boot the host from a SATADOM or disk device.

CPU
  • Sockets per host
  • Cores per socket
  • Number of vCPUs based on the expected number of virtual machines
  • vCPU-to-core ratio
  • 10% CPU overhead for vSAN

Host Networking

Provide more bandwidth for vSAN traffic to improve performance.

  • If you plan to use hosts that have 1-GbE adapters, dedicate adapters for vSAN only. For all-flash configurations, plan hosts that have dedicated or shared 10-GbE adapters.
  • If you plan to use 10-GbE adapters, they can be shared with other traffic types for both hybrid and all-flash configurations.
  • If a 10-GbE adapter is shared with other traffic types, use a vSphere Distributed Switch for vSAN traffic to isolate the traffic by using Network I/O Control and VLANs.
  • Create a team of physical adapters for vSAN traffic for redundancy.

Multiple Disk Groups

If the flash cache or storage controller stops responding, an entire disk group can fail. As a result, vSAN rebuilds all components for the failed disk group from another location in the cluster.

Use of multiple disk groups, with each disk group providing less capacity, provides the following benefits and disadvantages:

  • Benefits
    • Performance is improved because the datastore has more aggregated cache, and I/O operations are faster.
    • Risk of failure is spread among multiple disk groups.
    • If a disk group fails, vSAN rebuilds fewer components, so performance is improved.
  • Disadvantages
    • Costs are increased because two or more caching devices are required.
    • More memory is required to handle more disk groups.
    • Multiple storage controllers are required to reduce the risk of a single point of failure.

Drive Bays

For easy maintenance, consider hosts whose drive bays and PCIe slots are at the front of the server body.

Blade Servers and External Storage

The capacity of blade servers usually does not scale in a vSAN datastore because they have a limited number of disk slots. To extend the planned capacity of blade servers, use external storage enclosures. For information about the supported models of external storage enclosures, see VMware Compatibility Guide.

Hot Plug and Swap of Devices

Consider the storage controller passthrough mode support for easy hot plugging or replacement of magnetic disks and flash capacity devices on a host. If a controller works in RAID 0 mode, you must perform additional steps before the host can discover the new drive.