tikku@ginosko.UUCP (Sanjay Tikku) (04/26/89)
I am trying to do a comparative analysis of different UNIX (and look alikes) versions and determine which one is the most suitable as a platform for parallel processing work. I am thinking of either starting with SYSTEM V/BSD version and changing it suitably for multi-processor work or starting out with say Mach and then modify it appropriately to suit our needs here. If anyone has any opinion or comments on this subject it will be really helpful. I am specifically interested in knowing: a) Any literature which compares SYSTEM V and BSD versions regarding their adaptability for multi-processor work. b) Any literature on existing multi-processor systems like Mach compared to standard UNIX versions. c) If anyone has strong recommendations based on this kind of work done by them. Thanks in advance sanjay ================================+============================================= Sanjay Tikku | Voice: 508-685-7200 ext. 123 Sr. Software Engineer | FAX: 508-685-4940 SAMSUNG Software America, Inc. | Internet: tikku@samsung.com Computing Systems Division | UUCP: tikku@ginosko.UUCP One Corporate Drive | CSNET: samsung.com!tikku@Tektronix.CSNET Andover, MA 01810 | BITNET: samsung.com!tikku@Psuvax1.BITNET ================================+=============================================
oisin@cmtl01.uucp (Oisin) (05/12/89)
in Message-ID: <1073@altos86.UUCP> From: tikku@ginosko.UUCP > > I am trying to do a comparative analysis of different UNIX (and look alikes) > versions and determine which one is the most suitable as a platform for > parallel processing work. I am thinking of either starting with SYSTEM V/BSD I don't use the system, but I understand QNX may have much of what you are looking for. It is a very proprietary system, and I gave up trying to find out how compatible it is with standard UNIX(tm). One problem with QNX is that it runs on the Intel platforms only... i.e. XT, AT, and 80386 boxes. However, since it implements full distributed processing and real time operation, you can get a lot of work done on the better PCs. You can get information from Quantum in Ottawa, Canada. ........................................................................ Oisin "Curly" Curtin -- uucp:uunet!attcan!cmtl01!CN.YIS!oisin Disclaimer- Opinions are my own. CNR, AT&T and my wife don't want them. ........................................................................ Good judgement comes from experience... experience comes from bad judgement.
hynes@dg-rtp.dg.com (Ed Hynes) (06/12/89)
Take a look at DG/UX. It's (the only?) symatrical multi-processor and currently runs on two multi-processor architectures, our proprietary MV processor, and Motorolas' 88000 RISC processor. We're Sys V Verification Suite compliant, and committed to: SVID, BCS, OCS, POSIX, X, PEX (and probably some i don't know :-) It has many BSD extensions as well. Source licences are available. --- Edward B. Hynes | hynes%dg-rtp.dg.com@cs.relay.net Industry Standard Applications | hynes@dg-rtp.dg.com Data General Corporation | {backbone}!mcnc!rti!xyzzy!hynes 62 T.W. Alexander Dr. | R.T.P. NC 27709 | (919) 549-8421 fax 919 541 9313
boykin@encore.com (Joseph Boykin) (07/07/89)
In article <1212@altos86.UUCP> uunet!tps1.dg.com!dg-rtp!hynes (Ed Hynes) writes: > >Take a look at DG/UX. It's (the only?) symatrical multi-processor ^^^^^^^^^ I keep hearing folks from DG telling about their UNIX kernel, but I this posting I just *had* to respond to... No, DG's symmetrical multi-processor (both CPUS! (I know you're planning on 4, and maybe even 8)) is not even close to the only one available. Encore has been shipping symmetric shared memory multiprocessors (2 CPU minimum, 20 max; although the architecture supports 40). Our current product line is based on the National Semiconductor 32XXX series (NS32532 @ 30Mhz is ~ 8.5 MIPS). We're currently designing a Motorola 88K board (40 CPU's/system). We also have a DARPA funded contract to produce a 1,000 MIPS system, which will be based on multiple Multimax boxes using a total of 128 CPU's (again, all of these are symmetric shared memory multiprocessors). The following companies make multi-processor computers. Those with an asterisk (*) are symmetric multi-processors (the others may be, but I'm not sure). In alphabetical order: [Moderators Note:- You should add Altos to that list. They have a fully symmetrical multiprocessor 386 machine running 5.3.1 - Der] * Alliant (FX1/4/8/40/80) Amdahl (5890, 5990) Apollo (DN10000) * Ardent (Titan) Concurrent Computer (aka Masscomp) (Various models) Convex (C100, C200, C300) Cray (XMP, YMP) * DEC (3500, 6200, 6300, 8800) Edge (1000, 2000) Elxsi (6400) * Encore (Multimax) Flexible (Various) Floating Point (M64-60) IBM (3090 series) * Pyramid (Series 9000 * Sequent (Balance & Symmetry) Silicon Graphics (4D/200) Solbourne (?) * Stellar (GS1000) ---- Joe Boykin Encore Computer Corp Vice-Chair, IEEE Computer Societies' Technical Activities Board Internet: boykin@encore.com UUCP: encore!boykin
davidsen@sungod.crd.ge.com (William Davidsen) (07/12/89)
I think that Zenith has a multiprocessor 386 running a version of Xenix/386 or SCO3.2 which was shown at a show and should be shipping soon. I Havn't seen anything but the articles in the papers, anyone have more info? bill davidsen (davidsen@crdos1.crd.GE.COM) {uunet | philabs}!crdgw1!crdos1!davidsen "Stupidity, like virtue, is its own reward" -me
dworkin@Solbourne.COM (Dieter Muller) (07/12/89)
In article <3481@altos86.Altos.COM> apple!ames!EDDIE.MIT.EDU!encore!calliope!boykin (Joseph Boykin) writes: >The following companies make multi-processor computers. Those with an >asterisk (*) are symmetric multi-processors (the others may be, but I'm >not sure). In alphabetical order: > Solbourne (?) 'Fraid we aren't symmetric (yet). It's one master, N slaves. Dieter -- Welcome to the island. You are number six. boulder!stan!dworkin dworkin%stan@boulder.colorado.edu dworkin@solbourne.com