home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.parallel
- Path: sparky!uunet!think.com!rpi!gatech!hubcap!fpst
- From: dfk@dartmouth.edu (David Kotz)
- Subject: Re: mpp technology survey questions
- In-Reply-To: rms@well.sf.ca.us's message of Tue, 5 Jan 1993 19:36:14 GMT
- Message-ID: <DFK.93Jan6122807@wildcat.dartmouth.edu>
- Sender: news@dartvax.dartmouth.edu (The News Manager)
- Organization: Dartmouth College, Hanover, NH
- References: <1993Jan5.210849.26474@hubcap.clemson.edu>
- Date: Wed, 6 Jan 1993 17:28:07 GMT
- Approved: parallel@hubcap.clemson.edu
- Lines: 31
-
- In article <1993Jan5.210849.26474@hubcap.clemson.edu> rms@well.sf.ca.us (richard marlon stein) writes:
-
- Survey Questions on massively parallel computing technology
-
- I agree with Ian Kaplan in that Mr. Stein did not give sufficient
- context for this survey, and thus I would be a little dubious about
- answering it.
-
- I was especially surprised by the repeated contrast that seemed to be
- drawn between "scalable computation systems" and "shared-memory
- multiprocessor systems". Although I suppose some people might
- disagree, there is no particular reason why "shared-memory" systems
- cannot be "scalable computation systems". To put it another way, a
- scalable computation system might well be able to support shared
- memory. Look at the BBN Butterfly, KSR, and various software DSM
- systems and research prototypes like DASH.
-
- Most people who think shared memory does not scale are usually
- thinking that *bus-based* multiprocessors do not scale, which is
- basically true.
-
- Shared-memory multiprocessor <> Bus-based multiprocessor
-
- dave
-
- --
- -----------------
- Mathematics and Computer Science
- Dartmouth College, 6188 Bradley Hall, Hanover NH 03755-3551
- email: David.Kotz@Dartmouth.edu or dfk@cs.dartmouth.edu
-
-