which vv or host contributed to high service times?

[mohuddle]

Edit: We only have TWO CONTROLLER NODES

another view of just NL

[Richard Siemers]

Since you only have 2 nodes, a problem is that ALL of your NL capacity is on one cage/loop instead of distributed across all of your loops. You have 4 disk loops and all of your SSD sits on just 1.

[mohuddle]

I also opened a case with HP to do some analysis and wanted to share the following:

"...have found that you have oversubscribed NL drives and are experiencing Delayed ACK on node2 and node3. We recommend tuning some of your VVs that are currently on NL CPGs to FC CPGs. The delayed ACK will affect performance for all drive types as the system is waiting to return acknowledgment for NL drives. By tuning your NL VVs to FC…you should see the delayed ACK condition go away and your performance improve."

Utilizing the below, we will begin to find some of the culprits. "Listing of Vluns with high service times. Please tune some of your more busy VVs from NL CPGs to FC CPGs :"

Code: Select all

awk '$6>5||$12>20{print}' statvlun-rw-iter_60-d_30 | less
statvlun -rw -iter 60 -d 30

We've already tuned three DB Backup LUNs that were in an NL CPG to an FC CPG.