Xref: utzoo alt.sys.sun:3741 comp.parallel:2517 Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!stanford.edu!hubcap!fpst From: turner@sp1.csrd.uiuc.edu (Steve Turner) Newsgroups: alt.sys.sun,comp.parallel Subject: Re: Parallel computers from Sun? Message-ID: Date: 3 May 91 15:38:29 GMT Sender: fpst@hubcap.clemson.edu (Steve Stevenson) Reply-To: turner@csrd.uiuc.edu (Steve Turner) Organization: Center for Supercomputing R & D Lines: 69 Approved: parallel@hubcap.clemson.edu In-Reply-To: mok@fortsc.enet.dec.com's message of 1 May 91 18:54:00 GMT In article <1991May1.175839.25044@pa.dec.com> Charles P. Mok writes: > > I have heard some rumors (?) that Sun will be coming out with parallel > computers soon. We are talking about parallel processing, MIMD, not > just SMP and multiprocessors. Have anyone heard about this before? ... >From The New York Times, Sunday, April 28: : : Mr. McNealy will not say specifically when the company plans to : introduce high-end machines, but industry executives said it will : announce its first multiprocessor, code-named Galaxy, in June. But : special software that takes advantage of the machines' multiple chips : will not be introduced until the end of the year, when Sun unveils its : new operationg system, which will be usedd in Galaxy and was : originally developed with A.T.&T. Still just rumors, but (from my point of view, at least) more substantial ones. > Fro the uninitiated (myself), how do you think this kind of powerful > parallel offerings would affect the MIPS war in the workstation arena > today? What applications might the parallel box be suited for and > what would the conventional workstations be worthwhile for (in that > maybe not-too-distant future)? It remains to be seen how "powerful" Galaxy (and its ilk) will be, but no doubt HP, Next, Dec, and IBM have something parallel on the drawing boards. I find it significant that even the NYT finds it necessary to issue a caveat about about the availability of parallel software. This is the key, IMO. The applications that it will be suited for depend majorly on the method used to communicate between the processors. I would expect that initially a shared bus will be used and each processor will run separate processes, allowing only coarse-grain (but relatively easy to exploit) parallelism. This would be suitable for a machine that is intended as a compute server running a few to several CPU bound programs. To wit: : "We're not as strong as we should be in the computer server market," : said Mr. McNealy. "People have leapfrogged us, but we'll leapfrog : them again." Of course, there is software being developed at various places which will allow "automatic" parallelization of applications with internal parallism - and I doubt it's any coincidence that the folks I know who are graduating after doing work in that very area have recently been approached by "Sun Labs". All this is just rumor, naturally... -- Steve Turner (on the Si prairie - UIUC CSRD) ARPANET: turner@csrd.uiuc.edu Phone: (217) 244-7293 or (217) 367-0882 Kludges are conceived of man's natural fallibility, nourished by his loyalty to erroneous opinion, and perfected by the human capacity to apply maximum effort only when proceeding in the wrong direction. -- New Scientist 22 Dec. 1966 pg. 699 (from oed2, on-line OED) -- =========================== MODERATOR ============================== Steve Stevenson {steve,fpst}@hubcap.clemson.edu Department of Computer Science, comp.parallel Clemson University, Clemson, SC 29634-1906 (803)656-5880.mabell