> There's a NetBSD dialup server here that typically supports between
> 20 and 50 simultaneous users (It's a Pentium-100 with 42MB
> ram)... under version 1.0 the load with around 30 people logged on
> doing "average" stuff would be around 2... but now that the server
> is running 1.1ALPHA, the load is typically around 5 or 6, with no
> apparent slow down... I.E. response time is just as good. Any
> explainations?

It actually has 48MB of memory, and was running a fairly recent 1.0A
before we upgraded the kernel to 1.1 alpha.  I think the load is
typically higher because the machine can take higher loads before it
crashes.  (We're still getting crashes about once a day with "free
vnode isn't"; the vnode of /etc keeps showing up on the free list.  I
suspect that it's due to a bug in flock(), since we flock
/etc/master.passwd a lot.  We're still taking data right now.)

Sorry for the operational aside here.  If anyone does have any ideas
on why the load average might be artifically higher, I'm interested.