Hyper-V CSV datastore and VM physical disk latency

Question

I have noticed that the latency observed from perfmon counters inside a VM seems at times much higher than the latency for datastore observed from hyper-v host level.

I have a datastore that hosts 3 VMs, all of them are on one host.

The Physical disk performance object -> Avg. Disk sec/Read counter inside a VM at times show latency from 100 ms to 1000 ms in 1 hour period while CSVFS FS -> Avg. Disk sec/Read counter for the datastore at a host level where VMs are hosted shows values less than 100 ms.

The average read disk queue in the VM rises at that time to values in the range 50 to 350 and on the host to values in the range 40 to 150

Are the measurements right and is the situation normal? I could imagine latency being higher inside a VM if some IOs are kept in disk queue at the VM level and not passed to the host disk queue straight away, or maybe the IO from VM being throtled at host level, or VM threads being paused or context switched at the host level and the measurements regarding IO not taking it into account .

Is there anything that can be done tuning VM or host settings ?

Answer 1

Hi，

Thanks for your post! According to your description, seems the issue is Performance related, since Performance issue is complex to troubleshoot and measure, it's recommended to open a case with MS for better help.

Below is the link to open a case with MS:

https://support.microsoft.com/en-us/gp/customer-service-phone-numbers

Thanks for your time!
Best Regards,
Anne

-----------------------------

If the Answer is helpful, please click "Accept Answer" and upvote it.
Note: Please follow the steps in our documentation to enable e-mail notifications if you want to receive the related email notification for this thread.