Solved!
Go to SolutionW
wishtobeMrEMC
155 Posts
0
1793
November 13th, 2008 02:00
performance issue
we often get performance related critical alert associated with # of reads/sec and writes/sec. Below given is the details of reads and writes with memory cache size. Please kindly let me know what causing the error.
severity : critical
reads per sec : 1651
write % : 36
kbytes per sec : 66085
hits per sec : 2458
total hit % : 82
kbytes written per sec : 21135
throughput (kb/sec) 87220
writes per sec : 1176
i/o per sec : 2916
SRDFA max cache usage : 94
micro code : 5771.104
cache size : 128gb
severity : critical
reads per sec : 1651
write % : 36
kbytes per sec : 66085
hits per sec : 2458
total hit % : 82
kbytes written per sec : 21135
throughput (kb/sec) 87220
writes per sec : 1176
i/o per sec : 2916
SRDFA max cache usage : 94
micro code : 5771.104
cache size : 128gb
658
0
Posted November 13th, 2008 03:00
Are you talking about ECC alerts ??658
0
Posted November 13th, 2008 21:00
yes, Stefe658
0
Posted November 13th, 2008 22:00
I guess it's better to move this thread since ECC experts may shade more light on ECC alerts
Start counting down from 99 .. 98 .. 97 .. puff658
0
Posted November 13th, 2008 22:00
I was still counting .. and the thread found its better location .. I hope658
1
Posted November 14th, 2008 00:00
Hi,
All Symmetrix alerts in ControlCenter have pre-defined thresholds e.g. Front-end Fibre Channel Writes per sec have the following triggers:
- Critical >= 8000
- Warning >= 6000
- Minor >= 4000
There is no perfect number - they have to be analysed on a environment by environment basis. What is considered excessive at a small site may be quite normal at a large site.
It may be worth enagaging a local EMC performance specialist if you want to carry out a more granular analysis of any possible performance related issues.
Regards,
Séamus Coffey
EMC Global Services658
0
Posted November 17th, 2008 03:00
Hi bofh31337 and Seamus
thanks for your help. it was helpful. with the information given, let me figure it out the optimal threshold for performance in my env. thanks again.
cincystorage
2 Intern
•
467 Posts
658
0
Posted November 14th, 2008 06:00
Critical 8000+
Warning 4000-6000
Minor 0-4000
If performance in your environment is acceptable, what I would suggest you do, is use performance manager to collect a months worth of data. Then figure out average values for these numbers. That will give you a baseline of performance in your environment, and then use those numbers to alert on if you want.
Aside from the standard alerting values, performance manager collects a ton of statistics. What I like to do is define a whole bunch of new vital signs, not just the 15 default ones.
I break performance into a few categories. Then in each category, I look at several things:
Front end:
-Dir-Fibre>iops
-Dir-Fibre>% until
-Dir-Port>throughput in Kbytes per sec
-Dir-Port>% util
-Dir-Fibre>system write pending events
-Dir-Fibre>device write pending events
Cache/system:
System>total iops
System>% hit
System>% write
System>% read hit
System>Kbytes read per sec
System>Kbytes written per sec
Device>system bus Kbytes per sec
Dir-Fibre> % hit
Dir-Fibre>% write
Dir-Fibre>% read hit
Dir-Fibre>slot collisions per sec
Dir-Port>average io size in Kbytes
-System>system max wp limit
-System>number write pending tracks
-Dir-Fibre>system write pending events
-Dir-Fibre>device write pending events
-Device>max write pending threshold
-Device>write pending count
Back end:
-Dir-DA>iops
-Dir-DA>% util
-Disk>SCSI commands per sec
-Disk>% util
-Device>iops
-Device>HA Kbytes transferred per sec
-Device>DA Kbytes transferred per sec
-Dir-DA>prefetched tracks per sec
-Dir-DA>tracks not used per sec
-Device>prefetched tracks per sec
-Device>% read hit
Then I report on the states most critical and important to me. Bottom line is what numbers are good performance for you, are what you need to alert on.
Now that I¿ve finished my rant on performance, I¿ve got work to do