Re: Outstanding pre-production issues and next dev. cycle...

David S. Miller (davem@caip.rutgers.edu)
Tue, 27 Feb 1996 13:45:35 -0500


From: jon@gte.esi.us.es (Jon Tombs)
Date: Tue, 27 Feb 1996 17:08:29 +0100 (MET)

> The performance of the TCP code rivals if not clearly beats
> "production OS's" even on the same hardware. Of course, the is
> always room for improvement always ;-)

Performance is generally pretty good. The room for improvement is
probably to do fragmentation faster and without causing deadlocks
in the memory managment. 8k NFS will cause lockup a 4-8MB machine
as soon as the machine in under any combined NFS and memory load.

Right. I've found that playing around with the min_free_pages value
and similar can help avoid these situations.

Later,
David S. Miller
davem@caip.rutgers.edu