Our server(s) are dual socket 8core/16 thread Intel Dell’s. By VMWare’s own calculations that is 32 vCPU’s. Yet they make it difficult to setup a VM with the cores you want. Why is it broken down by socket and CPU with multiple ways to get there? For example, I want a 4core/8 thread VM running. why can’t i do this? I can only set it for 4 cores, either by 1 socket 4 cores, 2 socket 2 cores, or 4 sockets 1 core(why is this an option).
Can someone explain this easily so I can understand it and how do I determine which of those configurations is the better solution?
Edit: It’s 8/16 CPUS’s, apparently they are E5-2670’s
I’m guessing that is not really good, really high CPU Ready times? The green line is the CPU Ready time for one of the VM’s, it’s ready to go but the pCPU is just sitting there going … “I don’t have time!”
This is just from some googling and a basic understanding of how VMWare handles things. I think our server is over provisioned …
Some operating systems and software were designed with certain expectations and perform better in different scenarios. VMware gives you the flexibility to configure the VM so the software will behave optimally in the environment you give it.
Also the last document I read from vmware suggested to only provision up to the physical core count. Anything more would be over-provisioning and depending on the load could lead to slow downs. Even if something isn’t CPU intensive, simply occupying CPU cycles would deny other machines access to the CPU during that time (interrupts).
So in theory you should only have 8 vCPUs assigned at a time for optimal performance.
Not really understanding what you mean by that. VMware gives you the flexibility to assign cores however you… need… One socket 20 cores or 20 sockets 1 core. You’re expected to know what your server / application works best with.
