Recommendations for VM Resource Allocation

The allocation of resources for the Vocera Voice Server (i.e., processor, memory, etc.) for optimal performance depends on factors such as the volume of inbound data, processing overhead of the data type, and the complexity of the solution implemented.

Oversubscription

Oversubscription Type Description
Disc
  • No disk oversubscription is allowed on the host
  • All virtual disks should use a “thick” provisioning method
  • Only direct-attached storage is recommended
Note: If you are using NAS or SAN the maximum disk latency should be in line with the best practices for a real-time Voice application.

See VMware vSphere® 5.5 Documentation Center.

CPU
  • No CPU oversubscription is allowed on the host
  • The amount of provisioned vCPU's should be one less than the amount of physical processor cores available in the hypervisor (ESXi host)
  • CPU affinity is not required or recommended
  • Hyper-threading can be enabled but additional logical processors made available via the hyper-threading feature do not count toward the provisioning totals

Example:

Host machine contains 2 quad-core processors with hyper-threading enabled.

  • Physical CPUs = 8
  • Logical processors = 16
  • Maximum vCPU’s allowed = 7 (8 – 1 available for hypervisor use)
Memory
  • No memory oversubscription is allowed on the host
  • Provisioned memory of all virtual machines combined should equal 1GB less than the amount of memory installed in the hypervisor

Example:

  • Host machine contains 24 GB of RAM
  • 24 – 1 = 23 GB of RAM available for guest VMs memory allocation