Subject: Re: mb_map full and NMBCLUSTERS
To: Michael Graff <explorer@flame.org>
From: Andrew Gillham <gillhaa@ghost.whirlpool.com>
List: current-users
Date: 09/06/1996 14:21:00
Michael Graff wrote:
>
> Jason Thorpe <thorpej@nas.nasa.gov> writes:
>
> > I have mine bumped up to 1024 on the hp380 that is my ppp link and main
> > nfs server...
>
> Ok. kechara.flame.org gets:
> about 2000 hits/day on a WWW server
> has about 100 to 300 TCP connections open at once
> beats up on AFS a lot
> has 64M ram
>
> Do you think I should I go to 1024 or even higher?
Do some testing with ttcp, using udp and tcp. I ended up going to
*2048* on one machine because I still saw the error when flooding
with udp. Typically 512 should be fine, but if your machine has a
fair amount of network activity (sounds like it does) I would go with
1024 just to prevent it *ever* hosing up.
-Andrew