vSphere Latency Sensitivity and NSX Edge

Products

• VMware NSX

Issue/Introduction

Latency sensitive workloads may benefit from the vSphere setting "latency sensitivity" which is applied at the VM level.
It is not enabled by default on an NSX Edge VM but it is supported to configure this setting.

Resolution

1. Connect to the vSphere client.
2. Navigate to the Edge VM in the inventory.
3. Click Actions.
4. Click Edit Settings.
5. On the VM Options tab, click Advanced.
6. From the Latency Sensitivity drop-down menu, change the value from Normal to High.

Additional Information

Advantage of turning on latency sensitivity per VM :

Gives exclusive access to physical resources to avoid resource contention due to sharing.
With exclusive physical CPU (pCPU) access, each vCPU entirely owns a specific pCPU; no other vCPUs and threads
(including VMkernel I/O threads) are allowed to run on it. This achieves nearly zero readytime 

Once exclusive access to pCPUs is obtained, the feature allows the vCPUs to bypass the VMkernel’s CPU scheduling layer
and directly halt in the VM monitor (VMM) because there are no other contexts that need to be scheduled. This avoids the
cost of running the CPU scheduler code and switching between the VMkernel and VMM, leading to much faster vCPU
halt/wake-up operations.

Reference doc: https://www.vmware.com/docs/perf-latency-tuning-vsphere8