Subject: Re: load average (was: Re: How many sup sessions can lamp handle?)
To: Theo Deraadt <deraadt@sun-lamp.cs.berkeley.edu>
From: Havard Eidnes <Havard.Eidnes@runit.sintef.no>
List: current-users
Date: 11/10/1993 01:10:39
> the load does not go over 1.0, yet the subjective performance
> is terrible (likely because of interrupts)
> 
> With 12 max-connections, 40 disk irq's per second, and
> over 100 irq/sec ethernet. Squish, squish, squish.

I am of the opinion that the load average as computed on NetBSD is a wee
bit on the low side compared to other systems with similar loads, and I
have seldom (if ever) see it crawl above 1.0.  Can it be that processes
lying in "short-term disk wait" (or waiting for paging to complete) are not
taken into account when the load average is being calculated, or is there
some other lossage going on?  Or is it just me?  :-)

- Havard

------------------------------------------------------------------------------