From owner-freebsd-hackers  Tue Jul  8 04:18:28 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id EAA11761
          for hackers-outgoing; Tue, 8 Jul 1997 04:18:28 -0700 (PDT)
Received: from implode.root.com (implode.root.com [198.145.90.17])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id EAA11754
          for <freebsd-hackers@FreeBSD.ORG>; Tue, 8 Jul 1997 04:18:26 -0700 (PDT)
Received: from implode.root.com (localhost [127.0.0.1])
	by implode.root.com (8.8.5/8.8.5) with ESMTP id EAA08592;
	Tue, 8 Jul 1997 04:19:47 -0700 (PDT)
Message-Id: <199707081119.EAA08592@implode.root.com>
To: "Daniel O'Callaghan" <danny@panda.hilink.com.au>
cc: freebsd-hackers@FreeBSD.ORG
Subject: Re: news server behaviour 
In-reply-to: Your message of "Tue, 08 Jul 1997 17:45:58 +1000."
             <Pine.BSF.3.91.970708164958.869m@panda.hilink.com.au> 
From: David Greenman <dg@root.com>
Reply-To: dg@root.com
Date: Tue, 08 Jul 1997 04:19:47 -0700
Sender: owner-hackers@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

>730MB, sd1 is Quantum 1 GB. aha0 is 1542cf with 8Mbps scsi bus. 
...
>That all sounds reasonable, but I'm not so sure about the stats from top 
>and systat -vmstat.  Top shows innd (1.5.1) at 60-70% CPU (39% user, 50% 
>system).  Systat shows similar figures.
>
>I'm running kernel pppd with compression, and there seem to be about 4000 
>sio0 interrupts per systat refresh, but it shows 4% CPU for interrupt 
>servicing.
>
>/var/news is mounted noatime.  innd has 2 MB resident size.
>
>Is the CPU utilisation normal for the situation?  In a previous incarnation 
>the box was a Linux machine doing the same job under a different admin, 
>and CPU for innd was more like 5-20% with no readers.

   A couple of comments: FreeBSD might be accounting for the system CPU time
more accurately than Linux. I also see that you're using a 1542 with > 16MB
of RAM...which means that bounce buffers will be involved for about half of
the I/O. This could have a significant effect on system time when doing lots
of disk I/O. You didn't mention which version of FreeBSD you're using, but
you might want to try enabling the Pentium bcopy optimization if it's not
already enabled (make sure no npx0 flags are set). Is expire running? That
might help explain the high numbers.

-DG

David Greenman
Core-team/Principal Architect, The FreeBSD Project