johnw@astroatc.UUCP (John F. Wardale) (07/28/87)
Someone asked about "mainframe" vs "super-computer" In my mind, "super-computers" are good floating point, and are likely to be strange (specialized) beasts. A mainframe is a big, fast, reliable, machine. (Great at running drystones, and/or zillions of users) (Yes, a brand X computer could be a mainframe for one site, and a "super-computer" for another, but generally isn't) In article <10956@amdahl.amdahl.com> littauer@amdahl.UUCP (Tom Littauer) writes: >If the Dhrystone benchmarks are to be believed, ... >The May '87 ... Cray X-MP at 18,530, and the IBM 3090-200 at 31,250. and Cray-2=7480 and Amdal-5860=28,846 The Cray-2 SHOULD be close to the X-MP. A numericly intensive benchmark, LINPAK at n=300 has the following results: (All im Mega-FLOPS) Cray XMP-1 = 106; Cray-2 = 93; IBM 3090-200 = 18; (for comparison: DEC Vax11/780 = .11) This is why the "Crays" ar "supers" and the "IBMs" are "mainframes." Anyone have wetstone and/or other numbers for the above machines? > If you could get it, you'd want a machine fast enough to do EVERYTHING > quickly, not just a subset of things. Yes, but all these machine have engineering tradeoffs to "tune" them for the market nitch they were designed for. (This tuning starts in the Architacture, and continued thru the gate-level design.) John W - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Name: John F. Wardale UUCP: ... {seismo | harvard | ihnp4} !uwvax!astroatc!johnw arpa: astroatc!johnw@rsch.wisc.edu snail: 5800 Cottage Gr. Rd. ;;; Madison WI 53716 audio: 608-221-9001 eXt 110 To err is human, to really foul up world news requires the net!
eugene@pioneer.arpa (Eugene Miya N.) (07/29/87)
In article <373@astroatc.UUCP> johnw@astroatc.UUCP (John F. Wardale) writes: >Someone asked about "mainframe" vs "super-computer" > >In my mind, "super-computers" are good floating point, and are >likely to be strange (specialized) beasts. >A mainframe is a big, fast, reliable, machine. (Great at running >drystones, and/or zillions of users) >(Yes, a brand X computer could be a mainframe for one site, and a >"super-computer" for another, but generally isn't) Reliable? (well sort of...) and NO. Supercomputer is not a term relative to a site, it is relative to a period of time like now, or 20 years ago when a 7600 was a super. Only IBM would give you the brand X line. >In article <10956@amdahl.amdahl.com> littauer@amdahl.UUCP (Tom Littauer) writes: > >This is why the "Crays" ar "supers" and the "IBMs" are "mainframes." > >> If you could get it, you'd want a machine fast enough to do EVERYTHING >> quickly, not just a subset of things. > >Yes, but all these machine have engineering tradeoffs to "tune" >them for the market nitch they were designed for. (This tuning >starts in the Architacture, and continued thru the gate-level design.) One area where the IBM excell over the VAXen and SUN and where the Crays and Cybers excell over the IBMs is in the area of I/O. Basically (gross-oversimplification), VAXen are tied to the UNIBUS, MASSBUS, and SBI [yes yes BI, CI, etc.]. IBM has the "Channel." Crays and Cybers have propritary (until recently) channels of even higher bandwidth. These big machines also do disk stripping (Note: I have only seen two major papers on the performance of stripping, some PhD thesis are possible here). One of the reasons the Japanese machines are not popular is that they have few peripherals to things like Hyperchannel interfaces. (As well as batch operating systems: ACOS or more IBM flavored things.) I also have an interesting paper on IOStone from the guys who brought the massive memory VAX. Basically, I'll just read this discussion. You guys are on the right track. On to SIGGRAPH. P.S. I have asked one of the two men who defined the Class scheme of rating computers "Have we reached Class 7 with the ETA-10?" He's at LLNL. He said NO. I have to call the other one. (And Neil if you are reading this, George wants to talk to you.). From the Rock of Ages Home for Retired Hackers: --eugene miya NASA Ames Research Center eugene@ames-aurora.ARPA "You trust the `reply' command with all those different mailers out there?" "Send mail, avoid follow-ups. If enough, I'll summarize." {hplabs,hao,ihnp4,decwrl,allegra,tektronix,menlo70}!ames!aurora!eugene
chuck@amdahl.amdahl.com (Charles Simmons) (07/29/87)
In article <2418@ames.arpa> eugene@pioneer.UUCP (Eugene Miya N.) writes: >These big machines also do disk stripping > >P.S. I have asked one of the two men who defined the Class scheme of >rating computers "Have we reached Class 7 with the ETA-10?" > >--eugene miya > eugene@ames-aurora.ARPA OK. I'll bite. What is "disk stripping"? And what classes are defined and what do they mean? -- Chuck
greg@xios.XIOS.UUCP (Greg Franks) (08/10/87)
In article <11074@amdahl.amdahl.com> chuck@amdahl.UUCP (Charles Simmons) writes: >In article <2418@ames.arpa> eugene@pioneer.UUCP (Eugene Miya N.) writes: >>These big machines also do disk stripping >> >OK. I'll bite. What is "disk stripping"? Striping means assigning file system partitions across multiple devices so that contiguous file system operations can take place in parallel. Block 1 comes from disk 1, block 2 from disk 2 etc etc etc. There was a presentation at the Washington USENIX on "virtual disks" which briefly dealt with this topic. I'm sorry, but I can't supply any references. If somebody happens to know of any, send me mail, I'm interested too! -- Greg Franks XIOS Systems Corporation, 1600 Carling Avenue, (613) 725-5411 Ottawa, Ontario, Canada, K1Z 8R8 seismo!mnetor!dciem!nrcaer!xios!greg "Vermont ain't flat!"