Looking for GroupWise delivery statistics

Tom Miller tom.miller at ferc.fed.us
Thu Jul 1 12:57:10 UTC 1999


The Compaq 5000s were spiking CPU, so I moved the message processing POAs (there are six that do this) onto to Compaq 740.  That's a small server, but all it does is POA and nothing else.  It is a Pentium II 350 (?) and has 192 MB of RAM.  I was thinking of adding more RAM, but none of the server stats indicate more is needed (can't hurt though).

I know the POAs compete for the same resources, and that is probably one problem.  Yesterday I followed Novell's TIDs 2946201 and 2943356 and so far today I think messages are being processed much faster.

The indexing is done by the time users are coming to work.  Stangley, when the this occurs the CPU is not spiking and the server is otherwise behaiving normally.  

If this happens again then I'll know the changes I've made were not very helpful.

Thanks for the help
Tom

>>> "Tim O'Keeffe" <tokeeffe at frederickmd.com> 07/01 8:42 AM >>>
Tom

A couple of questions for you on your setups first:
are you running the servers using compaq's NIC's?  could be an issue but usually not.
how much memory and hard disk space do you have?  is it balanced as per novell's specs?
do you have your users connecting via mapped or ip connections?  mapped is much slower
do you have indexing running on the same POA's?  the frequency can slow the CPU's down.
are you getting the infamous 'CPU utilization errors' on the POA's?  this is really key.

i have about 600 users on each of our PO boxes,  they pump out about 53,000 messages a week.  it's not really a big load.  they are on compaq 1850R's with 512meg ram, with 18 gigs hard drives using about 12
gigs.  they used to be on 5000's with 512 megs ram and about 9 gigs hard disk space.

the 5000 did have a problem with CPU utilization, they were not optimized for gw, also did file and print services and three POA's running.  this box always seemed to run a little slower ??  we have other 5000's
and they seem to run fine.  never did track down the source of the pokiness.

we still have 3 poa's on the 1850R's and experience a little CPU utilization hiccups but not nearly as much.  one big difference is that we turned off indexing.  there is no document management running.

the next step for us is to monitor the server and see which process is grabbing the CPU and not releasing it.  haven't caught it yet.

let me know where you are.

tim o'keeffe
network engineer
cessi, inc


Tom Miller wrote:

> Can anyone tell me how many messages the POA, GWIA, MTA can process in an hour?  I recall seeing this somewhere but couldn't find it.
>
> I'm wondering because occationally some of our POAs have severe bottlenecks (about 300 user post office, Compaq 5000's, server optimized for GroupWise).  Even when I allocate more threads it just putts along.
>
> TIA
>
> Tom Miller
> FERC
> Tom.Miller at ferc.fed.us 





More information about the ngw mailing list