From owner-freebsd-smp Thu Jan 21 14:00:50 1999 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id OAA03949 for freebsd-smp-outgoing; Thu, 21 Jan 1999 14:00:50 -0800 (PST) (envelope-from owner-freebsd-smp@FreeBSD.ORG) Received: from oreo.adsu.bellsouth.com (oreo.adsu.bellsouth.com [205.152.173.36]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id OAA03944 for ; Thu, 21 Jan 1999 14:00:47 -0800 (PST) (envelope-from ck@oreo.adsu.bellsouth.com) Received: (from ck@localhost) by oreo.adsu.bellsouth.com (8.9.1/8.9.1) id RAA77667; Thu, 21 Jan 1999 17:00:09 -0500 (EST) (envelope-from ck) Date: Thu, 21 Jan 1999 17:00:09 -0500 From: Christian Kuhtz To: Licia Cc: smp@FreeBSD.ORG Subject: Re: smp and clustering Message-ID: <19990121170009.K5050@oreo.adsu.bellsouth.com> References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Mailer: Mutt 0.95i In-Reply-To: ; from Licia on Thu, Jan 21, 1999 at 01:06:41PM -0600 Sender: owner-freebsd-smp@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org On Thu, Jan 21, 1999 at 01:06:41PM -0600, Licia wrote: > > I'm becoming interested in the concept of clustering computers (at least > I believe that's the term for what I'm interested in :) ) and I was wondering > if someone could advise me as to the feasibility of modifying the SMP system > in FreeBSD to support sharing processors and memory with a completely separate > computer, over a high speed ethernet connection? Depends on how you define feasibility. What you're talking about is probably a distributed memory system, which is quite different from a shared memory system (SMP). AFAIK, from having thought about that a while back, it would be a (IMHO) fairly complex change, since FreeBSD wasn't designed from ground up to be that kind of box. When you have shared memory, you can "cheat" with interprocessor communication (because you got lots of it and it is very fast). Compared with distributed memory, where you have usually (among many other things) a fairly low speed internode link, you need to do things a little differently. You may want to experiment with PVM (and cohorts) if you are interested in distributed memory concepts. And you probably want to get familiar with MPI. Check out any of the well known supercomputing sites (or PVM's homepage for instance) for details. PVM, MPI and variants/related packages is was most people use these days when they talk about clusters (unless you got a very specialized machine). There's a very good book on clusters, btw.. Gregory F. Pfister In Search Of Clusters ISBN 0-13-899709-8 US$44.95 Cheers, Chris -- "Logic is a little bird, sitting in a tree; that smells *awful*." -- /usr/bin/fortune [Disclaimer: I speak for myself and my views are my own and not in any way to be construed as the views of BellSouth Corporation. ] To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-smp" in the body of the message