VM datastore Read / Write Latency

K

KapsZ28

2[H]4U
Joined
May 29, 2009
Messages
2,114
Is there an acceptable amount of latency for VM's with high usage? Below is a screenshot of the read/write latency to the datastore for a specific VM.



This is a Windows 2012 VM used for RDS with typically 25 people logged in. Traffic is lower right now than during the day. Is that kind of latency acceptable, or is this something you would work on improving and how would you improve?
 
Those max numbers are way out there.

Can you provide details on the hardware backing this?
 
The storage is a NetApp FAS3240 using NFS over multiple 1 Gbps links. It is a hybrid shelf with 18 - 900 GB, 10k, 2.5" SAS drives and 6 - 200 GB SSD. It is configured with 16 drives in RAID DP with 2 hot spares, 1 RAID group. The SSDs are used for Flash Pool. 5 SSDs in RAID DP with 1 hot spare. Only about 558 GB for Flash Pool.

Here is the latency for the volume on that NetApp over 24 hours. The majority of the latency on that volume is below 5ms.



The VM disk latency has about the same average of 20ms. I am guessing the maximums are so high from backups and deduplication running.

So I guess the real question is whether an average of 20ms disk latency is bad for a VM? Especially one used for RDS.
 
20ms might be just great. It might be terrible. Depends on the workload. Are the applications on the server suffering? Is performance a problem? As a general rule I don't like to see things sit above 20ms but peaks of 20ms are usually just fine.
 
Most everything is fine. Only occasional complaints from RDS users. This is not VDI and I think users expect the RDS environment to run just like they were using a desktop PC.
 
Hi NetJunkie,

Just wondering if you had any additional feedback based on the information I provided?

Thanks!
 
Honestly, given what you are running, I think these latency charts look just like what I would expect from that setup.
 
You must log in or register to reply here.
Back
Top