QoS Statistics and visibility

[frinku_hp]

I have QoS enabled Volume Set with min (5000), max iops (7000) and latency goals (10ms). The volumes are part of the AO policy and hence, auto-tiering to three tiers SSD, FC and NL. AO runs every night.

The QoS is working as expected like limiting the iops but not achieving the latency goals. Hosts sees high latency time. Investigating further I found out that iops limit is causing high service time on the hosts. Makes sense. According to the 3par architectiure, the iops are delayed at the controller level and sends the rejection if it sits there for more than 200ms.

I can monitor the rejections but no visibility of how much of iops intentionally delayed at the controller side. statqos doesn't display. Are there any way or method to see those values.

Thanks.

[Dennis Handly]

>Are there any way or method to see those values. 

srstatcmp mentions: delayed ack and dirty page

Does statvlun's -filt <fspec> with  <meas>  svct help?