rpcblast
Contributor
Contributor

ESXI CPU management/priority?

Here is my setup: Whitebox(well its black but thats beside the point) ESXI server, running a dual 1207 opteron nforce 3600 board(server boarD), 2x 2350 quad core optys, 32g ram, using ISCSI for storage. I am looking to set up one or two VMs to do . I know ehn you install , the CPU priority is set at idle, so that any other requests for CPU usage take precedence. How can I do this in vmware? Basically, I want the VM I set up with 4 vCPUs to only use whats not being used by any other VMs. I want it to be able to use as much as it can IF its available. so I dont want to set a fixed limit. I am running ESXI with virtual center.

0 Kudos
3 Replies
weinstein5
Immortal
Immortal

Will this be the only application running in the VM? IF it is then it will be running constantly and the vmkernel will schedule the vms virtual cpus on any available core - so it will only use what is available - but a couple of recommendations -

  1. If you were to use quad vcpus with 2 virtual machines I can qaurantee there will be contention and you would not be able to run additional vms on this box - this has to do you only have 8 cores and each virtua cpu will run on on a single core and one core will be used by the vmkernel itself -

  2. I would also reduce the number of virtual cpus in the vms - it easier for the vm to schedule a single cpu or even a dual vcpu vm than a quad

hope this helps -

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful
0 Kudos
rpcblast
Contributor
Contributor

Thanks for your quick response! I understand that the vkernal can dynamically allocate between cores. I have a total of 8 cores. and was planning on running 1 VM for folding with quad vcpu. What I am wanting however is that if the load gets high enough that other VMs need the additional resources, then the needs of the quad box will "take a back seat" until the load on the others calms down. I am running about 20 VMs total(most are low usage) and I only want this folding VM to make full use of the 4 cores IF nothing else wants them.

0 Kudos
weinstein5
Immortal
Immortal

The only way I see to do that is with cpu shares - that way you can prioritize which vm gets access to the resources when there is contention - so no contention the folding vm will get a whole core - with contention it will be to give some up - but I can not stress enough with a quad core vm on a host with 8 cores and 20 VMs I am almost positive I can gaurantee contention for cpu - I would recommend dropping the folding vm eith to a single vcpu or at a minimum a dual vcpu vm

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful
0 Kudos