From: Josh Guffin <guffin@purdue.edu>
Newsgroups: comp.parallel.mpi
Subject: Re: [Q] Diskless execution with MPICH ?
Date: Sun, 4 Jul 1999 11:22:22 -0500
Organization: Purdue University
Message-Id: <Pine.SOL.3.96.990704111845.27359A-100000@herald.cc.purdue.edu>
References: <377f5d7f.11406117@news.nic.utwente.nl>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
In-Reply-To: <377f5d7f.11406117@news.nic.utwente.nl>
Xref: ukc comp.parallel.mpi:5280


On Sun, 4 Jul 1999, Andrei A. Dergatchev wrote:
 
> Dear MPI experts,
> 
> I stumble with the following issue:
> From one hand, MPICH seems to require an executable
> to be present on each workstation in the cluster (I may
> be wrong with this though - I only recently started playing
> with included examples);
> From another hand, we want to avoid installing Linux
> on each workstation, so it would be preferable to boot
> them from network.
> 
> Is it possible to meet these 2 requirements simultaneously,
> or should I look for another package for distributed
> calculations ? Which one than ?
> 
> Thanks a lot for any hints,
> Rgds,
> 
> Andrei
> 
> 
Sure, it's possible. The diskless clients can have their root
directories NFS mounted from, say, /nfsmount/10.0.0.1 etc.  Their
programs will be located in, for example, /mpiprog, which corresponds
to /nfsmount/10.0.0.1/mpiprog on the BOOTP server (assuming that you
are using BOOTP).  Look for the HOWTOs on BOOTP, NFS, and Beowulf.

Good Luck.

Josh


=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
= Josh Guffin                                 guffin@purdue.edu =
= Purdue U. HEP - TASK E           expert.cc.purdue.edu/~guffin =
=                   #include <std/disclaimer>                   =
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=     

