High Utilization on GW Server

Mary Van Engelen maryve1 at starpower.net
Fri Dec 7 20:59:34 UTC 2001

We've been experiencing problems with slow response in GW.  The problem 
appears when there is high utilization on the email server where the post 
office resides. There doesn't seem to be a particular time of day when this 
happens, so I've ruled out indexing as the culprit.

Email Server: GW 5.5.4 non-EP with primary Domain/MTA/POA and an API 
gateway on a NW 4.2 server (280 mailboxes).
Gate Server: There's a secondary domain/MTA on another server that is 
running GWIA and WebAccess.

I did find that there was a high count (9,651) of "no ECB's available"  on 
the LAN card of the email server.  So last night I changed  several NW 4.2 
set parameters (minimum and maximum packet receive buffers, directory cache 
buffers and service processes, as well as timing parameters related to all 
three).  The changes I made were based on suggestions made in the October 
1999 Appnote for tuning, optimizing and sizing GW 5.5.

I've been watching it all day.  A few minutes ago I noticed a slow response 
while creating a message and ran to the server just in time to watch the 
number of directory cache buffers increase. High utilization stopped almost 
immediately. The "no ECB's available" count is still at zero.  I've seen no 
increase in packet receive buffers either.

So, what's the deal with the directory cache buffers? I'm off to the KB to 
do some research on it, but thought some of the great minds here might have 
some clues as to what I should do next.


More information about the ngw mailing list