Hello all,
I did a timing test of a DSMSERV UNLOADDB/LOADFORMAT/LOADDB with our 50GB
database on a separate system. This test system was an older, single CPU,
500MB RS/6000 and the dump was to a file on disk (approx 22GB).
What I'm surprised about is that when we ran the process on the real system
(8 CPU SP node, 2GB memory), it is no faster - in fact it might even be
slower (I'm not sure yet because it is STILL running from the weekend).
The only significant difference is that the real dump/load is using a 3590
tape instead of a file on disk.
This note is meant as a warning/heads-up to all of you, but I am also
interested to hear of other experiences/timings, and expecially from any
developer or support types with access to the source code as to why it
wouldn't be running faster on a faster system? Could using 3590 be the
problem? (Doesn't 3590 have good stop/start performance numbers?) I don't
suppose there are any "sleep 10" statements imbedded in the code for future
performance improvement releases ;-).
Thanks,
Scott
Scott McCambly
AIX/NetView/ADSM Specialist - Unopsys Inc. Ottawa, Ontario, Canada
(613)799-9269
|