Unsolved
This post is more than 5 years old
1 Message
0
410
April 15th, 2011 06:00
Symmetrix Perfromance issue with High Disk Utilisation
Hi,
I am hoping some of you good people can give me some pointers for resolving an intermittent performance issue I am having with a DX3000.
We have 4 MS Exchange Servers using several Devs (made up of 22 Meta's) on the Sym.
They have good performance most of the time.
However, When a report on another database is ran, Powerpath on the Exchange Servers show Q-IO's steadily go up on 4 devs on each Server and stay high. Other devs on the same server and same sym are ok have no Q-IO's.
When I check ECC Performance Manager, I can see Number of WP tracks reaches the System WP limit at the time of the problem, I can also see some disks have over 80% utilisation.
What I suspect is happening, is the other database has hypers on some of the same disks as the Exchange Servers and can't cope with the amount of IO, so the IO's are being held in Cache.
I am more used to working on Clariions, and if this was a Clariion I would just migrate the other database to a different Raid Group to see if this resolves the problem, Unfortunately for me this is a Sym where my experience is limited.
I understand Sym's do not have Raid Groups, disks are split in to Hypers, then devs are created on the Hypers.
What would be the best way to see if the database causing the issue is using hypers on some of the same physical disks as the Exchange Servers and how would I identify Devs that do not have hypers on these disks so I could move the problem database? Using the show dev command will give me the information but with so many meta's being involved this is not going to be easy.
I would appreciate any advice
Thanks in advance.


Quincy561
1.3K Posts
0
April 15th, 2011 08:00
The Symmetrix does have RAID groups, but they are not restricted to the whole disk, but are created at the logical volume level. Each logical volume can be considered a RAID group, similar to that on a CX.
You could try meta volumes, which would spread the load over more logicals, and therefore more RAID groups, but if there are already enough active, the system may be at the limit of what it can destage.
I would get a performance guru involved, and have them load your performance data into SymmMerge which could help you understand if it is a drive issue or a system issue.
dynamox
11 Legend
•
20.4K Posts
•
87.4K Points
0
April 16th, 2011 06:00
i assume you have EMC Control Center, if yes go to your array > Mapped devices > select appropriate type (IE: Meta RAID-5) . Right click on your meta > Relationship > Physical disks
SKT2
2 Intern
•
1.3K Posts
0
May 18th, 2011 18:00
I normally compare the disk groups with the RG on clar. If we decide to asign LUNs for exchange from a paricular disk group and dedicate another disk group for your other critical application that will provide an for the disks involved. This approach may have thier own pros and cons
symdisk -sid xxxx list -dskgrp_summary