Attachment uploads are currently disabled. This is a temporary situation and will resume as normal in the coming days.
UNSOLVED
D
Dhakshinamoorthy
2 Intern
•
174 Posts
0
881
December 8th, 2014 22:00
Difference in datastore level latency and backend LUN's response time
Hi There,
We have carved out a 500GB LUN from VNX5500 and presented it to 4 node ESXi cluster. Host team reported performance issues and they shared a screenshot which shows the datastore latency peaked to 71 ms. We enabled performance logging in VNX array and we analyzed the logs and nar file graph shows the LUN level latency(response time) is 42 ms.
I am bit confused, how there was a difference in datastore level latency and backend LUN's response time?
Regards,
Dhakshinamoorthy Balasubramanian
Dhakshinamoorthy
2 Intern
•
174 Posts
552
0
Posted December 10th, 2014 07:00
had a look in response time (ms)
Dhakshinamoorthy
2 Intern
•
174 Posts
552
0
Posted December 10th, 2014 08:00
around 20-25 minutes, latency value 71ms and 71ms+ in datastore side whereas maximum backend peak response time is 42ms.
those 71ms times ..are they picks or it actually stays consistently at those values. I am thinking it could be discrepancy due to polling interval differences where these picks do not translate properly into default 15 minute increments in analyzer. Just a guess.
is it something that is consistent and reproducible, you could lower "Real Time Interval" to 60 seconds and watch datastore performance real time. I would set the interval back to 300 seconds after you are done as it adds extra overhead.
dynamox
11 Legend
•
20419 Posts
•
87439 Points
552
0
Posted December 9th, 2014 06:00
when you looked in navi analyzer ..did you look are response time or service time ?