Re: Balancing disk drives

Ahhh making me say: "I Wonder" (and I don't have time at the moment to look it up) but do Journal Receivers get treated as other objects for purposes of balancing. Seems like a lot of the I/O in this environment is going to center on that. First writing them as data comes from the source and then reading them to apply it to the local DB.

Just thinking out loud at this point.

Also wondering if the fact that this is a very 'Write Heavy' workload if that matters to the balancing functions or the end result.

- Larry

On 7/29/2011 10:05 AM, rob@xxxxxxxxx wrote:

*This message was transferred with a trial version of CommuniGate(r) Pro*
I DID use the "add and balance". And it automatically did it twice.
Job log from job that started it:
Job 514520/ROB/DSP01 started on 07/26/11 at 14:03:53
QSYS. Job entered system on 07/26/11 at 14:03:53.
Message queue ROB is allocated to another job.
STRSST
wrkdsksts
signoff
Job 514520/ROB/DSP01 ended on 07/26/11 at 15:54:58;

DSPLOG
07/26/11-15:50:23: ASP balancing successfully started for ASP 1.
07/26/11-19:55:55: ASP balancing for ASP 1 ended.
07/26/11-19:55:55: ASP balancing successfully started for ASP 1.
07/26/11-20:23:38: ASP balancing for ASP 1 ended.

The ONLY other interactive type job during that time was Job
514488/PEGGY/GDSL621. While Peggy has that authority and is a great
operator when it comes to following a script she has a small comfort zone
and she would NEVER run STRSST or STRASPBAL and hasn't written CL in over
a decade. She get's on there just to get her BRMS stuff and that's it.

Job load is tough to measure. This a Mimix backup lpar for another lpar
on a different box in a different city. The same jobs are all that runs
"Did we get any journal entries, if so, post them". Reads are at a
minimum. No interactive, no batch. Where I really noticed the
performance was in the nightly BRMS backup. Added HOURS to the run.
Granted, the first night it was still balancing the disk drive. Backup
started at 17:30 and the balance didn't finish until 20:23. Following
night was still about 1.5 hours longer than prior to adding the disk arm.
After the *USAGE balance we're down to only an additional 20 minutes.


Rob Berendt