Occasionally, some administrative application, such as some audit software we have, will run simultaneously on several VMs. When this happens, there is a fairly noticable performance impact on all VMs. When looking at processes on VMs running the vmdesched.exe process, it's clear that some VMs are getting scheduled out (getting no CPU resources). However, when looking at CPU utilization on the ESX server, there are plenty of CPU resources available (see attached image). Any ideas why VMs getting scheduled out when there are plenty of PCPU resources? No other resources (RAM, network, disk, etc.) appear to be under any stress. Thanks.
There are no CPU limits defined and no affinities. ESX version is 3.5 Update 3. The VMs are all running Win XP SP2. The only known configuration 'issue' is that most of the VMs are using the Buslogic driver rather than the LSI driver. We are in the process of addressing this issue.
