Hi,
Because of limited resources and hardware maintenance i have had to run the majority of my VMs on a single host (vSphere 5.1) for a short period of time meaning that i am currently running approx 30 VMs with a vCPU count of 42 on a host that has 24 logical proccessors (2 x 6 core with hyperthreading). I had thought that generally its better to not go too much beyond a 1:1 ratio therefor keeping my vCpu count to around the 24 figure. Although perhaps i have this wrong as long as i monitor the CPU Ready time and this stays ideally somewhere under 5% or at least no greater than 10%?
Initally i was trying to measure the CPU ready time from the Hosts CPU ready time (converting from milliseconds to % using VMware KB: Converting between CPU summation and CPU % ready values). But have i misunderstood this and i should actually be checking the individual VMs CPU ready time and not using the figure for the host? I ask as i have some strange figures if i use the Hosts CPU ready time. e.g a max of 820000 Milliseconds over a week which would equate to a CPU ready time of 45.5%
If instead i use one of my most resource hungy servers that has 4 vCPUs and has had a maximum CPU ready time of 80000 milliseconds over a week so therefor a CPU ready of 4.44%. Understanding that this server will have to wait for all 4 pCPUs to become available then it should in theory be one of the worst hit servers and from these figures it would suggest that it would be comfortable to run in this configuration longer term?
It would be great if someone could clarify my questions around this and offer their thoughts on how much they overcommit CPU?
Many thanks
