Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!cs.utexas.edu!wuarchive!mit-eddie!uw-beaver!ubc-cs!alberta!aunro!atha!lyndon
From: lyndon@cs.AthabascaU.CA (Lyndon Nerenberg)
Newsgroups: comp.unix.ultrix
Subject: Poor NFS Performance (was Re: Managing a network of UNIX workstations)
Summary: Don't blame NFS
Message-ID: <1425@atha.AthabascaU.CA>
Date: 15 Jan 90 18:45:23 GMT
References: <3949@jhunix.HCF.JHU.EDU> <9338@cbmvax.commodore.com> <12938@watcgl.waterloo.edu>
Organization: Athabasca University
Lines: 25

In article <12938@watcgl.waterloo.edu> idallen@watcgl.waterloo.edu (Ian! D. Allen [CGL]) writes:
>We've had no success trying to do software development over NFS.  You
>can run applications over NFS, since all they have to do is load over
>the net once; but, the compile/edit/test cycle involves too many file
>accesses.  A compile or "make" can reference dozens of files, and the
>network overhead getting at each file was too much.  We gave it up and
>don't use NFS for much more than moving things from far away to local
>disk where we can work on it.

This sounds like a configuration problem, not an NFS problem. For the
last year I have been doing software development via NFS, first with
a VAXstation 2000 served by a Sun 3/160, and now with a SPARCstation I
served by a 3/160 and a 4/370. The VS2000 is not a fast machine in any
way, shape or form, yet I experienced very few problems with NFS. I have
compiled smail 3.1, TeX, large parts of the 4.3 source tree, etc., with
no difficulty.

Have you adjusted the retry, timeout, and write-size values for your
NFS mounts? Given enough wobbling of these parameters I have been able
to sucessfully use a 3b2 (!!) as an NFS server.
-- 
Lyndon Nerenberg  VE6BBM / Computing Services / Athabasca University
     {alberta,decwrl}!atha!lyndon || lyndon@cs.AthabascaU.CA

                     UREP: Peru in disguise?