I would like to hear some experiences others have had regarding setting
schedule randomization and maximum schedule sessions. What I am looking
for is the best compromise for best efficiency.

We currently are running a 6 hour wide start window for our nightly
backups with a maximum schedule sessions of 36. We back up approximately
74 nodes per night. This is working well for us most of the time. I am
looking to increase backup efficiency for those occasional nights where
a few nodes back up a larger than normal amount of data and the backups
run into daytime.  In the past we've used a lower figure for
randomization however on occasion, all the nodes clobbering the server
at once have caused the recovery log to become alarmingly close to
filling. (We are in Normal Mode.) A couple times the log filled up and
caused the operator system console com buffers to overflow. The log size
was increased as was the schedule randomization percentage to reduce the
parallel load on the server.

My question is would I gain efficiency by reducing the amount of
schedule sessions to the point where it doesn't stress the server and
drop randomization to 0? My thought is that I will gain maximum
efficiency this way since if I set max sched sess to say 25 there will
be always 25 sessions running in parallel and as soon as one completes,
another would immediately kick in rather than waiting a random amount of
time as set by sched. rand. pct.

Is my thinking correct and do you see a flaw in my logic?

 Thanks,
   Al

Alan Davenport
ALAN.DAVENPORT AT SELECTIVE DOT COM
Selective Insurance