Performance on MSA 1500

[Marcel Rond]

Hi,

We have a MSA 1500 with 3 disk enclosures.
We have 2 DL380 clusters running on it.
Cluster1 a/a : node1: exchange /node2: File.
Cluster2 a/p : SQL2005

Problem we have in the evening is that performance degrades for 20-30 mins.
Every evening at the same time.
It could be related to backup on HP MSL6000 Lib. connected to SAN. We backup the cluster through the san, and ik backs up other dl380 servers over the LAN.

When responsetime is slow, a server on the lan is backed up at that moment.

The computers experiencing the slow responses are accessing files on the Fileserver (on the Cluster)
Can someone suggest which steps I should take to pinpoint the source of the performance issue?
Could I best run perf. mon from microsoft?
any tips are welcome.

[James Muell]

Remember, ALL DATA I/O hitting the MSA platform goes through 1 controller rated @160mb/psec.

Not exactly clear about what's doing what. You say you are backing up the cluster (exch) via san backup. What about the other Fileserver cluster? Is that being backed up the same way?

Then you have other, san connected hosts that are being backed up over the LAN and that includes their mapped connections to the clustered fileserver?
I presume your not backing up the same data twice.
What about the Exch nightly utils? Are they running inside the backup window.

Are you running the MSa1500 Active-Active firmware, and distributing the luns across both controller?

[Marcel Rond]

Remember, ALL DATA I/O hitting the MSA platform goes through 1 controller rated @160mb/psec.

We have 2 clusters.
One cluster is active/active. With FileServer on one node and Exchange on the other node.

The second cluster is active/passive cluster with SQL2005 running on one node with one other node as failover.

These 2 cluster use the SAN for their storage.
Each Clusternode has it's own Arcserve backup job to backup data.

All the other servers are NOT connected to the SAN. These are connected by LAN.
One Server holds all backupjobs for these servers. These are then backed up through the LAN.

We are not backing up the data twice.

Are you running the MSa1500 Active-Active firmware, and distributing the luns across both controller? => I don't know. How can I check??

What about the Exch nightly utils? Are they running inside the backup window. => I don't know, how can I check?

Any suggestions to help me find where the bottleneck is?

[James Muell]

Your can check the MSA firmware by looking at the LED on the controller. If it is not immediately showing, use the up or down arrow buttons to step through until you see the firmware version listed. Older versions are 4.48, 4.94. Newer ones are 5.10, 5.20. All those are referred to as Active-Passive Controller firmware. There is new controller firmware that makes the 2 controllers Active-Active. It however was not MS Cluster certified the last time i looked.

Check with your Exchange admin to see what time the ESA utils run.

[James Muell]

Followup
You may need to correlate the scheduled time of:
Backups
Exchange Utilities activities
SQL DB processes, eg. backup or datadumps, batching, # of concurrent connections, data replication, log shipping.

A combination of any of these activities could be causing your 30 min degradation.
I'd be curious to know what you find.

[Marcel Rond]

Yes I think it might have to with multiple things running at the same time bringing down performance. Unfortunately i did not have time to investigate this. Hope I will have time soon.

I would like to add that the performance of the backup process is also strange.
We backup SQL-, Mail- and Fileserver (SAN)
When backing up Mail and SQL we reach around 2,5 MB/Sec
when backing up the Fileserver we reach only 0.7 MB/Sec, It takes almost 10 HRS to backup 420GB!!

When backing up the servers that backup through LAN instead of SAN we reach 1MB /Sec.
Even faster than the Fileserver which on the SAN.

Is fragmentation maybe causing perf. issues?
tnx