Operating System - OpenVMS
Showing results for 
Search instead for 
Did you mean: 

MSA1000 performance increase going from A/P to A/A fw ?


MSA1000 performance increase going from A/P to A/A fw ?

I have been asked to improve the performance of a MSA1000 in an OpenVMS 8.2 VMSCluster.

Current config.
Alpha ES47 one 2Gb HBA to external SAN SW Alpha DS40 one 2Gb HBA to same external SAN SW
External SAN SW has two fibre kables to the two ports on the MSA1000.

MSA1000 FW 4.48 dual controller each 256 Mb cache, 10 (active) disks 72 Gb 10K all RAID 1.

High IO count of size 4 and 16 blocks to very large files. ( the 16 block IO's is pt. 17 block IO's, but I will change that)

Both systems access disks, but main application runs on Alpha ES47. (failover cluster)
Options (random order)
1. Upgrade to firmware 7.0 and running A/A
2. Upgrade controller cache to 512 Mb each
3. Add 2 new HBA's and second SAN switch.
4. Upgrade the disks to 15K models.
5. Add MSA30-DB and move disks (assume pt. 4)
6. Add MSA30-DB and buy 10 extra disks.
7. Replace with MSA1500 + 2 MSA30-DB.
8. Upgrade to OpenVMS 8.3

I know some of this is free (FW 7.0, OpenVMS 8.3) and some is expensive (new 1500), but if cost is not an issue, how will you rank these in terms of performance improvement ?

How much would you expect/guess the performance will improve ?

Have you tried to run a MSA1000 A/A with OpenVMS. Are there any problems ?

Will extra cache be more needed with FW7.0 and A/A ?

What is the current status and best practise for disk IO size and alignment to MSA1000 Raid 1 volumes.
Is 2k 4k or other sizes best og is this only on EVA and Raid5 ?

Is there a way to monitor the performance of the MSA controllers?

Lars :-)
Honored Contributor

Re: MSA1000 performance increase going from A/P to A/A fw ?

Hi Lars,

1. As a first pass, I would be tempted to upgrade to the latest Active/Standby release for the MSA1000, 5.20. There are some performance benefits with that version over 4.48.

2. That would probably help.
3. That will help your resilience, but not necessarily the performance.
4. Would help.
5&6. May help, if you can increase the number of LUNs, and spread them over different controller ports.
7. Wouldn't make much difference. The MSA1500 and MSA1000 are esentially the same.

For performance monitoring on the MSA1000 you need to hook up the CLI port and use the command line options...


Esteemed Contributor

Re: MSA1000 performance increase going from A/P to A/A fw ?

hi lars,

rob already got you covered on the firmwares.

But your system performance may not depend on the entire MSA. The bottle neck must be in one or more of the LUNS.

If possible, post the following output:
$ monitor disk /per
$ sho dev d
$ sho dev/full 1$

From the MSA, attach serial cable which comes with msa and run the following command :
> show tech_support
> show units
> show disks

During msa is in use, especially where you encounter slow IO performance, at the MSA turn on the performance monitoring.
> start perf

Then extracts the info and capture this output too.
> show perf

After that, to stop the performance:
> stop perf

Post all the output. Let us know.


Honored Contributor

Re: MSA1000 performance increase going from A/P to A/A fw ?

What are you measuring to know if the performance has improved?
What is the criteria for improvement?
Throughput - Application work units done per second or application response time or something else?
Purely Personal Opinion