OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
Subject: Re: ideas on determining socket / OS latency
From: Nagendra Mishr (nboldfish.com)
Date: Mon Mar 20 2000 - 22:38:40 CST

Here's the latest...

We switched RedHAt 6.1 out for FreeBSD on one of the machines.
So now, M1 is RedHat 6.1 w/ postfix, and M3 is FreeBSD w/ postfix..

             , M1 , M2 , M3
Mar 17 18:00 , 538 , 0 , 3921
Mar 17 18:10 , 653 , 0 , 4243
Mar 17 18:20 , 171 , 0 , 65
Mar 17 18:30 , 13 , 0 , 5
Mar 17 18:40 , 0 , 0 , 0

That's the delivery's to the end users per 10 minute periods. M2 was
turned off.

The freeBSD machine is delivering almost 7 times as much as the redhat
machine..

In a previous run on the sending side, I did a Network Monitor capture
of the packets on the NT side... I found that there were 10 second
delays in which the sender was waiting for the result of the "." At
other times, the wait was insignificant.. I'll continue to capture the
packets to see where in the SMTP transaction the delay's are comming
from to see if there's a pattern.

I'm sure that there must be something in the way that RedHat is
configured that slows us down so much.

> The original poster's JAVA mail sending application pumps less than
> three 22kbyte messages/second into THREE Postfix servers, that's
> less than one message/second per Postfix server.

Clearly, not the case.. above.. more like 7 a second.. But those are
deliveries FROM postfix, not deliveries TO postfix.. I'm sure deliveries
TO postfix were higher.

Nagendra