Unsolved
This post is more than 5 years old
10 Posts
0
3559
Vmax Performance issue.....BE high utilization
The backend of 4 engine 4 bay 240GB cache Vmax is very very high and this frame is very high i/o frame. Is there any work around for better performance on this frame.
Quincy561
1.3K Posts
1
November 6th, 2013 09:00
Lower the workload (Host IO limits?)
Add more hardware
Change raid protection
Without more information, these are common solutions.
sanman999
10 Posts
0
November 6th, 2013 12:00
What work around can be done immediatly(temporary) to improve the performance. We did an SRDF to this frame and cant take performance questions from app team.
Quincy561
1.3K Posts
0
November 6th, 2013 14:00
Not sure what the performance problem is, or if it even has anything to do with the backend or not.
I would get a Symm performance person involved. There should be local folks that can assist.
Host IO limits could slow less critical loads if the workload is just too much.
Fenglin1
2 Intern
2 Intern
•
2.1K Posts
0
November 7th, 2013 00:00
Are you able to identify which host consume much resource? Comparing the IOps and MBs/sec, greate than 100KB IO occurr from you APP server. I suspect there are Apps has big I/O which utilize much backend resource. Maybe writing lot of data which cause Cache WP% is high.
I'd like to indentify the hosts and apply hostIO limiatation function may help temperary and follow Quincy56's suggestion to engage EMC local team for next steps
apethe
34 Posts
0
November 8th, 2013 09:00
I am having the same problem. BE utilization and response time is high. There is no SRDF. Please suggest a fix.
Quincy561
1.3K Posts
0
November 8th, 2013 10:00
it can be difficult and time consuming to try and solve performance problems here. I would suggest a CS performance case be opened with EMC. It sounds like there is too much workload going to too few SATA drives from what you have described.
apethe
34 Posts
0
November 8th, 2013 10:00
microcode is 5876.229.145. the application team is complaining of degraded performance since we migrated to the vmax. they have response times around 200ms for the sata disk group.
as for the 3 'fixes':
host io limits are set as per emc recomendations
we cannot add more hardware
sata is at r6
Quincy561
1.3K Posts
0
November 8th, 2013 10:00
Apethe, what code is the box running?
With 5876.229 the DA % busy may be reporting higher than what should be reported.
My first post suggested 3 "fixes".
apethe
34 Posts
0
November 8th, 2013 11:00
i am seeing the same load on all sata disks
Quincy561
1.3K Posts
0
November 8th, 2013 11:00
Just because it is the same load to all drives doesn't mean it isn't too much.
Figure each disk can do 50 IOPs with good response time. Each host write = 6 disk IOs. So you can figure out how many SATA disks you need pretty easily for your specific workload.
SreeHari_Karana
36 Posts
0
November 10th, 2013 22:00
Hi,
Do you use virtual provisioning in your environment? And how are the metas configured? Striped or concatenated? What is the predominant RAID type?
Based on your answers for the above, I can may be suggest a few things, but my suggestions would really be more of the best practices, I'd recommend you engage a perf specialist from EMC.
Thanks,
Sreehari
Quincy561
1.3K Posts
1
November 11th, 2013 06:00
Apethe,
Are all drive types RAID 6? If so, that would explain the high backend utilization.
apethe
34 Posts
0
November 11th, 2013 06:00
Hi sreehari,
The entire storage is virtual provisioned
We have 8 way to 32 way metas, striped
Raid type is R6
Ameya
apethe
34 Posts
0
November 11th, 2013 07:00
Yes, all the drive types in the screenshot are raid 6. Even with FAST working optimally, the server teams are complaining of high response times.
Quincy561
1.3K Posts
0
November 11th, 2013 08:00
There is nothing wrong with using RAID6 that adding more drives and/or DAs won't fix.
However your performance issue could be something else. Again, this should be dealt with via the SR process along with the local performance folks.