ADSM-L

expiration bottleneck

2001-11-05 09:24:18
Subject: expiration bottleneck
From: "Richard L. Rhodes" <rhodesr AT FIRSTENERGYCORP DOT COM>
Date: Mon, 5 Nov 2001 09:20:40 -5
YASEPQ - yet another slow expiration processing question!!!!

Our expiration processing currently takes around 4 hours and
returns similar results each day:

        Examined:       7,558,141
        Expired:        197,717

Specs:
        db size:        56gb
        utilized:       76%  (around 40gb)
        buffpool:       1gb
        cache hit:      97.7%

        dbbackup takes around 1hr.

What is greatly confusing us is the sar statistics
from during expiration.  Below is a small sample, but the
entire 4hr run looks just like it:

16:39:23       2       2       7      90
16:40:23       3       1       6      90
16:41:23       3       1       6      90
16:42:23       2       2       6      90
16:43:23       3       1       6      90
16:44:23       3       2       6      90
16:45:23       3       1       5      90
16:46:23       4       1       5      89
16:47:23       2       1       7      90
16:48:23       5       3       5      88
16:49:23       4       2       3      90
16:50:23       5       2       3      91
16:51:23       5       2       3      90

Quite simply, the system isn't being pushed at all.  The system
has lots of processor available and it's not in a large wio state.

The disk subsystem is a IBM Shark with the db spread across
2 raidsets (16 drives).

Q)  What can we do to speed expiration up?

My only thought is to spread the db across disks, but since I'm not
is a high wio state, I don't think that will help.

Any thoughts are welcome!

Thanks

Rick
<Prev in Thread] Current Thread [Next in Thread>
  • expiration bottleneck, Richard L. Rhodes <=