VM = Windows 2k8r2 standard, 8gb mem, 2 sockets x 3 cores
The new VM/app is some kind of network monitoring tool
CPU % Util is in the 90's (10-12ghz)
Wait ms is in 8k
Ready % is less than 1
Co-stop per CPU is +150
Mem is 60-75% util
Host:
2s x 10core
Ready % less than 1%
CPU Util is less than 25% - was less than 15% before new VM
Co-stop jumped from near 0 to 50's with new VM
11 other guests of varying configs, no apparent issues.
No complaints (yet). Are there any other metrics from host or VM to keep an eye on.
I would be concerned about adding more vCPUs to the detriment of the host and other guests through increased co-stops.
Any thoughts?
If you've got 6 vCPU sized properly for the NUMA node running ~90% util, it sounds like it's sized about right.
Watch the CPU Contention % for the VM and Host to detect CPU pressure. Also, keep an eye on your vCPU to pCPU density.. most people run 3:1-5:1 depending on the SLA or tier of workload. Between watching the CPU Demand %, CPU Contention %, and CPU density ratio you should be able to increase vCPUs on the hosts while ensuring you're not negatively impacting your existing VMs.
Thanks!
VM CPU contention is around 4%
Host contention is just under 2%
Host CPU demand jumped from around 10% to nearly 40% when this app came online
vCPU to pCPU ration is only 1.7
This is certainly one of our heavier hitters, and worth keeping an eye on.
Hi
Welcome to the communities.
Generally now day most of the higher version application need high resources.
so nothing to be worry , but if you are planning to upgrade you need to be very very careful.
