From owner-freebsd-stable  Mon Dec  1 09:36:47 1997
Return-Path: <owner-freebsd-stable>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id JAA24526
          for stable-outgoing; Mon, 1 Dec 1997 09:36:47 -0800 (PST)
          (envelope-from owner-freebsd-stable)
Received: from pop.uniserve.com (pop.uniserve.com [204.244.156.3])
          by hub.freebsd.org (8.8.7/8.8.7) with SMTP id JAA24520
          for <FreeBSD-stable@freebsd.org>; Mon, 1 Dec 1997 09:36:44 -0800 (PST)
          (envelope-from tom@uniserve.com)
Received: from shell.uniserve.com [204.244.210.252] 
	by pop.uniserve.com with smtp (Exim 1.73 #1)
	id 0xcZjl-0006jz-00; Mon, 1 Dec 1997 09:34:25 -0800
Date: Mon, 1 Dec 1997 09:34:24 -0800 (PST)
From: Tom <tom@uniserve.com>
To: David Greenman <dg@root.com>
cc: Mattias Pantzare <pantzer@ludd.luth.se>,
        Justen Stepka <jstepka@chaos.winternet.com>,
        FreeBSD-stable@freebsd.org
Subject: Re: CPU Load 
In-Reply-To: <199712010829.AAA02884@implode.root.com>
Message-ID: <Pine.BSF.3.96.971201092138.28657A-100000@shell.uniserve.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-freebsd-stable@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk


On Mon, 1 Dec 1997, David Greenman wrote:

> >On Mon, 1 Dec 1997, Mattias Pantzare wrote:
> >
> >> On Sat, 29 Nov 1997, Justen Stepka wrote:
> >> 
> >> > Recently I added memory to my NFS server (dx4-100 now w/ 32 megs of RAM),
> >> > when I did this the overall system preformance increased dramiticly. The
> >> > problem that I noticed was that when using NFS/NIS the CPU load climbs to
> >> > about 4.0+, is there a special reason that this might be happening?
> >> 
> >> The load value is not CPU load. It is the average number of processes
> >> ready to run or waiting for disk I/O to complete.
> >
> >  No, it is the average number of processes that are ready-to-run.
> >Processes waiting for disk io (or any io) are not ready to run.
> 
>    Mattias Pantzare is correct. The load average is the number of runnable
> processes plus the number of processes in a short-term (disk) wait. It's
> meant to indicate the overall effect on interactive users, not necessarily
> just the use of the CPU.

  Short term disk wait?  Is there a long term disk wait too?  What is the
difference?

  There seem to be some differences in how different Unix systems
calculate the load average.  Linux seems to do something quite different
for one.

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

Tom