In practice the MPU rating of a virtual Avaya Aura® MS will vary widely depending on the performance of the underlying hardware. Providing a single MPU rating for a virtual appliance is not possible unless the hypervisor hardware is fixed, or the capacity is aligned with the least performing hardware. Therefore, our estimates are conservative in some cases, depending on the actual hardware Avaya Aura® MS is deployed on.

Due to the fact Avaya Aura® MS is a low latency real-time application, guidelines must be followed when deploying an Avaya Aura® MS virtual appliance on a shared hypervisor. Oversubscribing the hypervisor can result in unpredictable performance due to varying VM scheduling latency. An ordinary application may not be impacted by this latency, but a real-time application needing to send and receive media at very frequent and precise intervals (<= 10ms) is at greater risk. For these reasons, the total vCPUs utilized by all virtual machines on a shared hypervisor should not exceed the number of logical cores available. This prevents virtual machines from interfering with each other, and prevents difficulties with scheduling each VM consistently and predictably.

For example, a hypervisor with two E5-2640v3 CPUs will have 16 physical cores total, or 32 logical processors with hyperthreading. If an AAMS VM is deployed on this hypervisor, a total of 32 vCPUs are available for assignment across all co-resident VMs including AAMS. Assigning more than 32 vCPUs will oversubscribe the server. Also note that although it seems capacity is being left on the table, HT provides only a modest improvement in performance (10-15%). The improvement is potentially zero or negative in virtualized environments with latency sensitive, CPU intensive applications. For more information, please consult the AAMS documentation.