Improving Stun Times

[Tony_Rosas]

VMware 6.0

Compellent SC5020 SAN connected with 10GB

VM Size 1.6TB

VM Workload: SQL

12 disks split among 3 paravirtual controllers

OS disk using LSI Logic SAS

Does anyone know if there is anything else we can do to increase performance for stun times? We are getting really good stun times but sometimes the stun times go up.

I pulled the stun times for the last 30 days for this VM and for the most part it's a tenth of a second or half a second but sometimes it increases to 1.5 seconds with the highest being at 1.9 seconds.

The average for the month is .59 seconds, I know these stun times are probably already in the "excellent" category but I would like to see if there are any tweaks to bring them down even further. especially the ones that creep up over a second.

I attached our report of stun times for this VM for the last 30 days in case anyone's interested.

[scott28tt]

Moderator: Moved to vSphere Storage

[daphnissov]

Stun times are heavily linked to underlying storage performance and that of the hypervisor as well, so unless you have some of that other data and have attempted some correlation, just a table of those times isn't enough to go on. And on that point, it would be better if you could paste that text in-line to your response (VMTN has an insert table feature). It's generally not safe for strangers on the Interwebs to be opening MS Office documents from other strangers.

I would begin by looking at your Compellent array and correlate the stun times with those metrics to see if you're getting a spike at the storage layer or what that storage says at all.

[Tony_Rosas]

Thank you for your response.

My next step is to work with Dell Compellent support and see if we can correlate those times and IO performance. Just wanted to show the stun times list is already good and you are correct, on it's own they mean nothing. Hopefully it will spark a conversation with anyone that was trying to improve their environment as well.

[Tony_Rosas]

I worked with Compellent support and they were unable to suggest anything either. The IO burst that happens during a snapshot is so insignificant and seems they only suggest following the best practices guide, which we already have.

[glaucoporzio]

What is the VMFS version used in your case?

thanks

G

[Tony_Rosas]

Currently running VMFS 5.