Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!wuarchive!gem.mps.ohio-state.edu!lavaca.uh.edu!uhnix1!sugar!ficc!peter
From: peter@ficc.uu.net (Peter da Silva)
Newsgroups: comp.arch
Subject: Re: parallel systems
Message-ID: <6655@ficc.uu.net>
Date: 24 Oct 89 14:25:40 GMT
References: <35825@lll-winken.LLNL.GOV> <20336@princeton.Princeton.EDU> <7651@bunny.GTE.COM> <1989Oct23.152120.25967@cs.rochester.edu>
Reply-To: peter@ficc.uu.net (Peter da Silva)
Organization: Xenix Support, FICC
Lines: 20

In article <1989Oct23.152120.25967@cs.rochester.edu> crowl@snipe.cs.rochester.edu (Lawrence Crowl) writes:
> SHARED MEMORY HAS A COST.  Implementing shared memory over a scalable
> interconnect may require a larger aggregate bandwidth than that of distributed
> memory systems..  I don't think there has been enough research here to know
> the real tradeoff, but such a result would not suprise me.

Also, the job of programming a shared memory system is a lot harder than
programming a system with messages as the communication medium. The number
of successful message-based operating systems demonstrates this.

Of course you can implement messages in shared memory (a trivial proof is
aforementioned operating systems), and gain a perfromance improvement. The
question is whether this offsets the extra bandwidth.

Ah well, I'd much rather have both.
-- 
Peter da Silva, *NIX support guy @ Ferranti International Controls Corporation.
Biz: peter@ficc.uu.net, +1 713 274 5180. Fun: peter@sugar.hackercorp.com. `-_-'
"That particular mistake will not be repeated.  There are plenty of        'U`
 mistakes left that have not yet been used." -- Andy Tanenbaum (ast@cs.vu.nl)