Industrial vSwitch (IvS) supports realtime applications that requires low latency packet forwarding. To realize the low latency forwarding, NUMA settings are needed. Without it, IvS could face the high latency issue.
This article describe how to determine the correct NUMA configuration is set for IvS.
4.2.x
All the VMs must have NUMA alignment with the physical NICs and NSX Lcores.
For identifying physical NIC NUMA alignment, use the following ESXi command on the host.
[root@localhost:~] vsish -e get /net/pNics/vmnic2/properties |grep NUMA
Device NUMA Node:1
On NSX, make sure that Transport Node Profile has a CPU config with the same NUMA index as NICs for Host Switch. If the CPU only has one socket, the default NUMA node index is 0.
For VM config, set numa.nodeAffinity the same NUMA index from vCenter UI for the list of vPLCs VMs. This parameter is not necessary if the CPU only has one socket.