Update on Mail on AIX and System Performance/Emergency Downtime

As of 8:30 AM this morning, oboe (the AIX cluster mail server), managed 
to process between 4,000 and 6,000 mail messages in a couple of hours.  
This is a radical improvement in performance based on two changes made to 
oboe.

The first and biggest change was in moving the entire password data set to
oboe and not letting oboe use YP in resolving email addresses (use of YP
resulted in a problem whenever we received a message with a bad To:
address).  This change is responsible for the biggest part of the change 
in the way oboe has been handling mail resulting in a 90% decrease in the 
backlog on the local AIX clients in that two hour period.

The second change, which was temporary, was to take oboe off the Internet 
and just let it communicate with the local cluster.  This was to give it 
a chance to deal with local mail only and has been reversed as of 8:30 
this morning and Internet mail from other hosts is again being 
delivered.  Oboe's current load is due to it playing catch-up with these 
mail sources.

***DOWNTIME***

IBM is loaning Cal Poly 256 MB of memory for oboe to improve its 
performance in handling mail.  This additional memory will allow oboe to 
be able to deal with larger quantities of mail while maintaining a more 
stable operational level.

To install this memory, oboe will be taken down at 3:00 PM today and will 
be down until slightly after 5:00 PM.  This time is required to install 
and test the additional memory before beinging oboe back on-line.

PERFORMANCE

The additional memory combined with this mornings change should allow 
users to see a more stable environment as oboe heads toward a stable 
operational level.  Users should see far fewer long delays and fewer NFS 
errors in the cluster.

Thank you all for your patience!