2016-07-26 11:51 PM
Tell me please, how can I perceive situation when there is a high (about 100 ms) latency shown in the OnCommand System Manager console while no workload is on the storage. I understand that this latency is not related to the production IO workload and disk latency, but why does OnCommand System Manager shows the latency? What processes are there on the storage system and where does these values come from?
2016-07-31 11:48 PM
1. in some extrem situation (like yours, no workload at all), the latency number is skewed due to no enough sample (coz all the numbers are average)
2. another possibility is that system's doing house-keeping job (like disk scrub, volume scrub etc)
2016-08-01 12:38 AM
Thank you, Yaoguang!
You are absolutely right, but I need little more explanation about the "no enough sample coz all the numbers are average".
Does it mean that monitoring system polls the storage system and receives any random number as a latency value (in particular, 100 ms that is quite a lot even for house-keeping job)? Or does the storage responds that latency is zero (or about zero) but monitoring system interpretes it as 100 ms?
Clarify this point please.
2016-08-01 08:38 PM
ok... just for example... ur measuring time is 5 min, the value you're looking is nfs reading latency. so all the nfs reading requests finished in this 5 min will be counted. let's say 2 nfs reading requests finished in this 5 min. they are 20ms and 25ms. then the nfs reading latency is 22.5ms which is the number you see. in some situation, there's no any request completed in ur sample time, then your denominator to calculate ur average is 0, so you see the skewed number.
if still not clear, please let me know