MSA1000

[Lawrence-]

Hi Guys,
I wonder if I can pick your brains regarding a msa1000 san I've inherited recently, It's been sitting for a number of years and not been maintained / touched.

Currently im trying to understand where the potential bottlenecks in this system may be - There are two msa 1000 dual controllers running in an active / passive setup, there are a number of servers linked via 2 32 port brocade switches.

My main concern currently is our file/print cluster - the troughput we're seeing averages at 1mb/sec on a good day we'll see a burst upto 8-10mb/sec.

When monitoring the file and print cluster - i see that the memory pages/sec is averaging at around 360 with a maximum of 1210 and the avg. disk queue length is averaging at a 4.418 and a maximum of 29.496.

Read/write on the msa controllers is set to 50/50% - expland and rebuild priorities are set to low, surface scan delay is set to 3.

On the fpc - each drive on the san has write caching and advanced performance enabled.

Would anyone be able to point me in the right direction as to where to look for potential bottlenecks? Or is this normal for this model?

Another question I have - would defragmenting the drives improve performance? I suspect there has been no maintainance on this system since it was installed in 2004.

[Víctor Cespón · Accepted solution]

First thing would be to update the firware to lastest version (5.30).

The performance depends mainly on how many disks there are on the MSA. 10 MB/s is not a problem, unless it's done in 4 KB random transfers. You should see how many IOPS the servers perform. That's usually where the limit is.

Yes, defragmentation will reduce the number of random jumps on the drives, giving a faster access time.

[Johan Guldmyr]

Hi, I don't know what is the normal speeds/values with the MSA. Depends on many things, see below for my initial thouhts :)

Active/Active Firmware is available for the MSA1000 with a later firmware. That might be able to speed things up for you.

Might be that the disks themselves as well are the cause of the low performance, maybe they/some of them need to be replaced, or maybe firmware updates on them.

Is it only one MSA1000? No extra disk enclosures?

How about the block size on the volumes?
Have you looked into if that can be optimized?

If you hook up a serial cable to the msa1000 you can run a "show tech_support" to get the status of the MSA.

With that you'll see what kind of cache modules you have isntalled. It is possible to get a cache upgrade to 512MB on the MSA1000.

On the SAN - how is the zoning done?
Are the ports set to fixed or autonegotiate?
Fixed is usually recommended.

See the quickspecs:
http://h18004.www1.hp.com/products/quickspecs/12624_div/12624_div.html

MSA1000 firmwares:
http://h20000.www2.hp.com/bizsupport/TechSupport/SoftwareIndex.jsp?lang=en&cc=us&prodNameId=305215&prodTypeId=12169&prodSeriesId=377751&swLang=8&taskId=135&swEnvOID=1005

Quite a lot needs to be done to change from A/P to A/A, so read up on that.
Multipathing/MPIO, DSM, zoning, etc, probably needs to be configured.

[Lawrence-]

We're testing the io of our servers.

Have also made noises about purchasing the 256mb upgrade.

Plan to upgrade firmware when we can arrange downtime.

The both of the MSA1000 have 2 extra enclosures.

At the moment we have the ports set to autonegotiate - as we're having trouble setting the emulex cards to 2g (HBAnyware 4.1.1.35 detects no cards)

Ironicaly, adding dns and domain info to one of the brocade switches (it wasnt there previously) seems to have improved performance.

I've done a show tech_support and got a face full - which is attached - Can anyone see anything glaringly obvious here?