what does M&R show for response times on those LUNs ?

This could be due to a number of things. It's hard to say for sure without actually looking at nar files.

Is there low throughput (IOPS) on these LUNs? The number of IOPS are used in calculating response times so if there are low amounts of throughput, this can sometimes result in erroneous response times on LUNs in Unisphere Analyzer and M&R.

Are you using Windows 2012? This could be a potential issue.

Are you using iSCSI? This could be an issue if iSCSI is used.

Those are just a couple of ideas but are both worth looking at.

Thank you Dynamox, Joshua,

The luns are showing 60-100ms. This is a Linux server using native MPIO.

Admingirl

how many IOPS per LUN do you see ?

Hello,

I let this go for a few days to see if the issue originally reported was just a one day thing, however, it's not. The number of IOPS for these luns is only 100 per. These are Oracle RAC database luns servicing year end processing.

Thanks again!

Admingirl

ok, how about the drives that are members of the pool where these LUNs reside, how many IOPS do you see on average ?

Hello Admingirl,

I agree with Joshua about response time measurement not being reliable in case of low I/O, and here's some further explanation for why this is:

- Response Time is calculated using the I/O queue length divided by the throughput. So, as the throughput gets lower and approaches zero, the Response Time can show some very high values, which do not indicate any performance bottlenecks necessarily.

- This is documented in this Knowledge Base article: https://support.emc.com/kb/174073.

I would take the response time measurements on the storage array seriously in case there is significant throughput on the LUN being analyzed, or if the host side is experiencing this latency or showing it in a host side performance monitoring tool as well.

To check for performance bottlenecks overall, try, for example, seeing how much throughput storage pool disks are under (and whether the IOPS per disk are under the threshold for each disk type), Storage Processor write cache levels or Storage Processor utilization.

Also helpful in regards to high LUN response times is this KB article: https://support.emc.com/kb/91353. Although this may not be as relevant, since you already confirmed low queue depth and service time. Still, its useful article .

If you have any questions, just let me know.

Adham Makady

VNX Customer Services

Hi Joshua,

thanks for pointing this out.

In my view this is massive and needs flashing lights...

182688 : VNX: Storage Processor (SP) response times can increase when Windows 2012 issues Trim/Unmap commands to thin LUNs.            

https://support.emc.com/kb/182688

Kind regards