I saw some reference architectures that shows the vcpu/mem requirements, but I would like to see if someone saw some specification that shows the ratio between vcu and sockets. I mean for example if I am using the medium size related to vRA reference architecture how many vcpu/sockets I will need to configure for each IaaS servers.
Yes, it's a general VMware best practice to use sockets over cores. From a vCPU perspective, both result in the same number, but it allows better NUMA scheduling at the hypervisor layer. There are dozens of articles and books that cover this, so some basic searches will turn up lots.
vRA components in the stack should (like is generally recommended) prefer sockets over cores.
Hi daphnissov, thanks a lot for your reply, I saw the minimum and maximum requirements in the vRA reference architect docs that only mentioned as you mentioned recomendation, but nothing saying that we must be use sockets over cores, is it a VMware best practices? Do you have some link that mentioned this assumption?
Yes, it's a general VMware best practice to use sockets over cores. From a vCPU perspective, both result in the same number, but it allows better NUMA scheduling at the hypervisor layer. There are dozens of articles and books that cover this, so some basic searches will turn up lots.
Thanks, I will look for this topic...I appreciated a lot.