stein@dhw68k.cts.com (Rick 'Transputer' Stein) (11/18/89)
The New York Times reported today that E&S is bowing out of the
superconfuser business. The aritcle, by John Markoff, states that
their system, believed to be a VLIW flavor, is having both hardware
and software production problems, and that the performance is not
quite as competitive as they had once thought. I guess those
Livermore Loops are worth something!
--
Richard M. Stein (aka, Rick 'Transputer' Stein)
Sole proprietor of Rick's Software Toxic Waste Dump and Kitty Litter Co.
"You build 'em, we bury 'em." uucp: ...{spsd, zardoz, felix}!dhw68k!stein brooks@vette.llnl.gov (Eugene Brooks) (11/20/89)
In article <27611@dhw68k.cts.com> stein@dhw68k.cts.com (Rick 'Transputer' Stein) writes: >The New York Times reported today that E&S is bowing out of the >superconfuser business. E&S did not see the Killer Micros coming. Any vendor who invested time and money in a "custom" CPU implementation over the past several years is getting eaten alive by the sudden onslaught of the Killer Micros. You can't sell an expensive slow computer in a competitive market. Its going to be a terrible year for vendors of custom architectures, new vendors will be completely flushed and old vendors will barely survive on the hysterisis of their existing customer base. Even the CEO of Cray Research mentioned the RISC microprocessors in the recent Supercomputing '89 conference in his keynote address. He referred to the performance increases of microprocessors as "astounding." brooks@maddog.llnl.gov, brooks@maddog.uucp
thomson@cs.utah.edu (Rich Thomson) (11/20/89)
In article <38966@lll-winken.LLNL.GOV> Eugene Brooks writes: ]In article <27611@dhw68k.cts.com> Rick 'Transputer' Stein writes: ]> The New York Times reported today that E&S is bowing out of the ]> superconfuser business. ] E&S did not see the Killer Micros coming. Any vendor who ] invested time and money in a "custom" CPU implementation ] over the past several years is getting eaten alive by the sudden ] onslaught of the Killer Micros. You can't sell an expensive ] slow computer in a competitive market. Wether or not E&S saw the "killer micros" has nothing to do with why they shut down their supercomputer project. Hardware problems coupled with the expense of such a project (especially for a company the size of E&S) were the major driving factor. If you take the time to look at the ES-1 (the name of the product) architecture you will see that it is basically "killer micros" all connected together through shared memory running Mach. It is NOT a vector optimized machine. The fact that the "killer micros" keep getting more powerful says something for custom design. What do you think they make the microprocessors out of? Surely not gate arrays. ] Its going to be a terrible year for vendors of custom architectures, ] new vendors will be completely flushed and old vendors will barely ] survive on the hysterisis of their existing customer base. That depends on what your custom architecture is; building a custom architecture doesn't necessarily mean that somebody with 5 micros under their arm is going to blow your pants off. -- Rich Rich Thomson thomson@cs.utah.edu {bellcore,hplabs,uunet}!utah-cs!thomson "Tyranny, like hell, is not easily conquered; yet we have this consolation with us, that the harder the conflict, the more glorious the triumph. What we obtain too cheap, we esteem too lightly." Thomas Paine, _The Crisis_, Dec. 23rd, 1776
brooks@maddog.llnl.gov (Eugene Brooks) (11/20/89)
If your basic processor is not as fast as a current killer micro, someone with ONE Killer Micro will blow your pants off. This was the case in an extreme for the ES-1 and is why Evans and Sutherland saw no hope at all for the future and closed shop. One can quote the offical words printed in the press, but the bottom line is that you can't make money selling an expensive slow computer. Had there been any money at the end of the tunnel, they would have hung in there and solved their hardware and software problems. The Killer Micros were moving in the on the territory and will have completely dominated it before they could get their problems fixed. It took the use of on the order of 10 processors (processing elements) or more on the ES-1 to match traditional supercomputer performance. As noted in an earlier post, the ES-1 was a nice micro architecture, but without any of the Killer part in either performance or cost. Judging from the Livermore Loops figures for the latest Killer Micro from hell, the MIPS R6000, it matches traditional supercomputer performance with ONE processor, and not just for "scalar only" codes. The other micro vendors will soon follow with even more terrible critters, its a competitive world after all. Killer Micros are certainly custom architectures which cost a lot of money to develop. The difference for them is that their development costs are amortized over large markets and the parts end up sold at "cookie cutter" costs. Compare this to the costs of developing the Cray-3, mentioned by Rollwagen at SC'89 to be more than a hundred million dollars. This development cost has to be amortized over the sale base, and for a market which might only be few tens of machines (after SSI, Cray Computers, Cray Research, Tera, Convex, and the three Japanese companies divide up the total market which the Killer Micros leave for them) puts quite a lower bound on the machine price before you even get around to charging the cost of the hardware. This is in sharp contrast to the situation for the Cray-1, the sale of the first copy of which more than paid the entire development cost for the machine. Gone are the days of high profit margins for supercomputers. Perhaps I am wrong and the R6000 powered box should be sold for more than a million and MIPS is just dumping it on the market to destroy the supercomputer vendors. I don't see any "anti dumping" legislation in the works, however. One thing is clear, if traditional supercomputers don't find another order of magnitude in single CPU performance real soon, at fixed cost, they will not survive The Attack of the Killer Micros. For scalar codes supercomputers need even more leverage, two orders of magnitude. I don't think it is going to happen. brooks@maddog.llnl.gov, brooks@maddog.uucp
mash@mips.COM (John Mashey) (11/20/89)
In article <38980@lll-winken.LLNL.GOV> brooks@maddog.llnl.gov (Eugene Brooks) writes: .... >Perhaps I am wrong and the R6000 powered box should be sold for >more than a million and MIPS is just dumping it on the market to >destroy the supercomputer vendors..... Just so they're no confusion: it shouldn't be, and nobody's dumping... -- -john mashey DISCLAIMER: <generic disclaimer, I speak for me only, etc> UUCP: {ames,decwrl,prls,pyramid}!mips!mash OR mash@mips.com DDD: 408-991-0253 or 408-720-1700, x253 USPS: MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086
rodman@mfci.UUCP (Paul Rodman) (11/20/89)
In article <38980@lll-winken.LLNL.GOV> brooks@maddog.llnl.gov (Eugene
Brooks) writes: ....[ deleted soapboxing about Killer Micros]....
Ok, ok, so I'm wasting my time, but I just can't let Mr. Brooks
continue his deluge of propaganda without throwing in my opinions....
Let's ignore the fact that the Killer Micros usually don't have a
decent memory system or I/O system. [Not to mention that their
compilers and O/S are often not very robust...:-)]
There IS a grain of truth in what Mr Brooks says, in that I do belive
that there has been, and will continue to be both a factory cost and
performance compression from the lowest to the highest performance
computers.
Today's dense CMOS and ECL ASICS appear from PCs to workstations to
supercomputers, as do pretty dense packaging techniques. So it IS
getting harder to figure out how to apply current technology to build
a supercomputer, i.e. how do I use more hardware to build a faster
uniprocessor machine?
<Commercial ON.>
One technology exists that CAN use more hardware for more performance:
the VLIW machine. Furthermore, such a machine uses replicated
functional units, lots of SRAM, and minimal control, all of which
contribute to ease of the design cycle. The ease of use of the VLIW
system [compile-and-go] is important for the Non-Brookses of the
world, belive me.
I have seen the state of typical software efforts in the good 'ol USA
and I feel that it will be a long time before efforts such as that of
Mr Brooks are the norm. [100's, 1000's of micros used for
time-to-solution]. God knows enough folks are working on, and have
worked on, the problem! As such multiprocessing techiniques improve,
they WILL be used commercially, but mostly with more modest numbers of
VLIW "supercomputers".
I DO agree with Mr. Brooks that single cpus built with 100's of
different ASIC designs, dozens of boards, and mediocre performance, are
dead meat in the not-so-long run [e.g. Cyber 2000]. But Mr. Brooks
errs in thinking that single-chip cpus are the be-all, end-all of cpu
design.
<Commercial Off.>
Speaking as a engineer that does NOT work on Killer Micros, I just
don't seem to feel as depressed about the future of high-$$$ cpu
design as he thinks I should.... :-) Mr Brooks' attitude reflects his
rigid thinking, and his underestimation of change in areas other than
the one he has focused on. His refrain reminds me of all the wags 10
years ago that claimed "The Attack of the Killer CMOS" would remove
LSI ECL from use in computer design. Instead almost all high-end cpus
sold today are ECL.
Paul K. Rodman / KA1ZA / rodman@multiflow.com
Multiflow Computer, Inc. Tel. 203 488 6090 x 236
Branford, Ct. 06405
rcd@ico.isc.com (Dick Dunn) (11/23/89)
In article <1128@m3.mfci.UUCP>, rodman@mfci.UUCP (Paul Rodman) writes: >...Let's ignore the fact that the Killer Micros usually don't have a > decent memory system or I/O system... No, let's ignore it because it's not a fact but an incredibly biased (and not particularly useful) opinion. (Actually, what I've seen suggests that the memory systems are usually pretty well balanced, and that the I/O systems are out of balance to about the same extent they are on most machines...I/O has been in catch-up mode for years.) >...[Not to mention that their > compilers and O/S are often not very robust...:-)] Yeah, let's not mention that, since it isn't true. -- Dick Dunn rcd@ico.isc.com uucp: {ncar,nbires}!ico!rcd (303)449-2870 ...`Just say no' to mindless dogma.
mcdonald@aries.uiuc.edu (Doug McDonald) (11/23/89)
In article <1989Nov22.175128.24910@ico.isc.com> rcd@ico.isc.com (Dick Dunn) writes: >In article <1128@m3.mfci.UUCP>, rodman@mfci.UUCP (Paul Rodman) writes: >>...Let's ignore the fact that the Killer Micros usually don't have a >> decent memory system or I/O system... > > >(Actually, what I've seen suggests that the memory systems are usually >pretty well balanced, and that the I/O systems are out of balance to about >the same extent they are on most machines...I/O has been in catch-up mode >for years.) > I contend that there is no such thing as an "out of balance IO system". Certainly there is for memory vs. cpu. But there is an extremely wide range of needed ratios for io vs cpu power. Certain business uses need vast IO compared to CPU, some scientific uses need 99.999% cpu and .001% IO. There is an obvious needed market for a powerful CPU with far less IO power than a Cray or an IBM mainframe. One person's "balance" is anothers overkill - both ways, of course. Doug McDonald
seanf@sco.COM (Sean Fagan) (11/24/89)
In article <1989Nov22.175128.24910@ico.isc.com> rcd@ico.isc.com (Dick Dunn) writes: >In article <1128@m3.mfci.UUCP>, rodman@mfci.UUCP (Paul Rodman) writes: >>...Let's ignore the fact that the Killer Micros usually don't have a >> decent memory system or I/O system... > >No, let's ignore it because it's not a fact but an incredibly biased >(and not particularly useful) opinion. What?! Uhm, I hate to tell you this, but the reason mini's provide much better throughput than micros was because of the I/O subsystem. A '286 is faster than a VAX (785, let's say), but, go multi-user, and there is no comparison: the VAX will get *much* better throughput (meaning: it may take three times as long to do your sieve of Erasthoneses, but swapping processes in and out, and doing any disk I/O, is going to go more than three times faster). This is because the '286 (a killer-micro, compared to a VAX 8-)) doesn't have the memory or I/O subsystems that the VAX does. Now, compare just about *any* system on the market today with, say, a CDC Cyber running NOS (or, deity forgive me, even NOS/VE). (Bet you were all waiting for me to mention those, weren't you? 8-).) As has been pointed out before, lots of people don't *need* 250 MFLOPS / MIPS on their desktop; they just need to shuffle data back and forth (that's why there's a TPS [Transactions Per Second] measurement; any commentary on that, John? Michael?). Without a decent I/O subsystem, you won't be able to do this. And the memory in most "killer micros" is defficient because I can't do *real* DMA (it tends to steal cycles from the CPU). (N.B.: some K.M.'s *do* have *real* DMA. I'm waiting for them to come out with *real* I/O subsystems [using, say, a 68000 as a PP]. Then they will scream, even compared to a Cyber.) >>...[Not to mention that their >> compilers and O/S are often not very robust...:-)] >Yeah, let's not mention that, since it isn't true. Some of the compilers available today are pretty amazing, especially compared to what was available just a decade ago. The OS's running on most K.M.'s, however, tend to be unix varients (or, deity help us all, DOS). This is not a terribly robust OS, nor a terribly quick one (asynchronous I/O would be really nice; there are some other things that could be useful). So, yeah, it is true. -- Sean Eric Fagan | "Time has little to do with infinity and jelly donuts." seanf@sco.COM | -- Thomas Magnum (Tom Selleck), _Magnum, P.I._ (408) 458-1422 | Any opinions expressed are my own, not my employers'.
brooks@maddog.llnl.gov (Eugene Brooks) (11/24/89)
In article <3893@scolex.sco.COM> seanf@sco.COM (Sean Fagan) writes: >for me to mention those, weren't you? 8-).) As has been pointed out before, >lots of people don't *need* 250 MFLOPS / MIPS on their desktop; they just >need to shuffle data back and forth (that's why there's a TPS [Transactions >Per Second] measurement; any commentary on that, John? Michael?). Without No one needs a computer of any performance level on his desk. What one needs is a modern windowing terminal on a desk, connected to the computer in the computer room with a connection of suitable bandwidth to handle the drawing on the screen. The Killer Micros and striped disk farm belong in the computer room where fan noise and heat does not bother anyone. A Killer Micro on ones desk is just a waste of a Killer Micro, along with a uselessly small main memory size. The utilization of such a machine is so low it is hard to measure reliably. >most "killer micros" is defficient because I can't do *real* DMA (it tends >to steal cycles from the CPU). (N.B.: some K.M.'s *do* have *real* DMA. >I'm waiting for them to come out with *real* I/O subsystems [using, say, a >68000 as a PP]. Then they will scream, even compared to a Cyber.) A 68000 is probably not fast enought to handle IO for a good killer micro. A real computer will have handful of Killer Micros hooked up to a coherent cache system, with possibly VME DMA IO on the main bus or some adapter attached to it. A supercomputer will have a scalable coherent cache system and some number of these caches hooked to striped disk farms to supply the serious IO needs of such a machine. Don't confuse the basic CPU technology of Killer Micros with the really poor main memory and IO systems which people sell as single user workstations. >Some of the compilers available today are pretty amazing, especially >compared to what was available just a decade ago. The OS's running on most >K.M.'s, however, tend to be unix varients (or, deity help us all, DOS). >This is not a terribly robust OS, nor a terribly quick one (asynchronous I/O >would be really nice; there are some other things that could be useful). Killer Micros will soon dominate the world of computing, UNIX already does. DOS users are not computing, but saying just what they are doing is not appropriate for public consumption. brooks@maddog.llnl.gov, brooks@maddog.uucp
iyengar@grad1.cis.upenn.edu (Anand Iyengar) (11/24/89)
In article <39361@lll-winken.LLNL.GOV> brooks@maddog.llnl.gov (Eugene Brooks) writes: >No one needs a computer of any performance level on his desk. What one >needs is a modern windowing terminal on a desk, connected to the >computer in the computer room with a connection of suitable bandwidth to >handle the drawing on the screen. Interesting statement. I'm running off of an X term now, and while it's not bad, it's not trouble free. If the serving host, the network, or the terminal itself are down, I can't use it work get work done. If I have a high-performance micro with a reasonable drive, I only crash when the local machine/site has problems (let's not argue NFS; that's somewhere in between, depending on how much you mount and keep locally. Diskless clients have many of the same problems as windowing terminals). Also, performance of the Xterm decreases with loading of the central host, and network. It's not blazingly fast in itself, either. Doing anything really "graphical" on it bogs it down or crashes it. Forget animation or really neatsy stuff. Maybe one could put a faster CPU, and internal bus in it to get the graphics to go fast. But then why not just go the extra 5 yards, and drop some more RAM and a drive on it, and make it a real-live computer? >The Killer Micros and striped disk farm belong in the computer room where fan >noise and heat does not bother anyone. Why? I hate the noise as much as anyone, but why is it bad to have a high-performance computer on your desk? Drop a net link to it, and people can log in to it from about anywhere. >A Killer Micro on ones desk is just a waste of a Killer Micro, along with >a uselessly small main memory size. The utilization of such a machine is so >low it is hard to measure reliably. It is and it isn't. Price/performance doesn't scale linearly. It's not clear that a big mainframe is lots (your mileage will vary) better than a number of micros. There are still some things that I can't do on a mainframe that I can do on a micro, such as crash it. Because we have a number of small boxes around people can just connect to a different one for a while, and it's not a problem. >In article <3893@scolex.sco.COM> seanf@sco.COM (Sean Fagan) writes: >>K.M.'s, however, tend to be unix varients (or, deity help us all, DOS). >>This is not a terribly robust OS, nor a terribly quick one (asynchronous I/O >>would be really nice; there are some other things that could be useful). Agreed, but people are band-aiding as they go, and it's wide-spread enough that it will probably be here a while. >Killer Micros will soon dominate the world of computing, UNIX already does. >DOS users are not computing, but saying just what they are doing is not >appropriate for public consumption. In every DOS user is a potential UNIX user. You might not like DOS, but that doesn't make it evil. Funny; IBM use to think the same thing... Anand. -- "I've got more important things to waste my time on." {arpa | bit}net: iyengar@eniac.seas.upenn.edu uucp: !$ | uunet --- Lbh guvax znlor vg'yy ybbx orggre ebg-guvegrrarg? ---
seanf@sco.COM (Sean Fagan) (11/25/89)
In article <39361@lll-winken.LLNL.GOV> brooks@maddog.llnl.gov (Eugene Brooks) writes: >In article <3893@scolex.sco.COM> seanf@sco.COM (Sean Fagan) writes: >>most "killer micros" is defficient because I can't do *real* DMA (it tends >>to steal cycles from the CPU). (N.B.: some K.M.'s *do* have *real* DMA. >>I'm waiting for them to come out with *real* I/O subsystems [using, say, a >>68000 as a PP]. Then they will scream, even compared to a Cyber.) >A 68000 is probably not fast enought to handle IO for a good killer >micro. A PP-type processor does not need to be fast, really. If it's fast enough, you turn your system into a dual-processor system, with heterogeneous processor types (it can be done, and has been. Mach can, I think, be made to work rather well with it). By having a, say, 16-MHz 68k serve as the I/O processor for a KM (say, a 67MHz R6k), and doing the system correctly, then the 68k still has a bit of idle time (say, 1-5%, not counting time spent waiting for i/o to complete). More, and you should probably retune / redesign your system; less, and you should have a slightly faster processor. PP's for a Cyber are *slow*. But they get the job done real well. >Killer Micros will soon dominate the world of computing, UNIX already does. I don't think so. IBM still dominates the world of computing, along with FORTRAN and COBOL. Personal computers are catching up, though. Give it another 5 or 6 years (i.e., more people use an IBM mainframe than use an PC [except, possibly, as a terminal to the mainframe]). >DOS users are not computing, but saying just what they are doing is not >appropriate for public consumption. They're using computers, aren't they? Guess what they're doing, then: they're computing. A very small percentage of computer users need pure number-crunching power (or else everyone would go out and buy a Cray or i860 8-)); a larger number of users would like to see more MIPS (for drawing speed) and more *throughput*. Again, as I've said before, a CDC Cyber 170/760 is slower, MIPS-wise, than quite a few of the newer RISC systems out there. However, it *feels* faster because of the throughput difference, even with 100 users on it (speaking from experience). When you have a system that can compile a 10 000-line FORTRAN program in less than 40 seconds, *without* going through a cache, then I'll be happy with a KM. Until then, however, the mainframes are going to win, and continue to be bought. -- Sean Eric Fagan | "Time has little to do with infinity and jelly donuts." seanf@sco.COM | -- Thomas Magnum (Tom Selleck), _Magnum, P.I._ (408) 458-1422 | Any opinions expressed are my own, not my employers'.
bzs@world.std.com (Barry Shein) (11/26/89)
>I don't think so. IBM still dominates the world of computing, along with >FORTRAN and COBOL. Personal computers are catching up, though. Give it >another 5 or 6 years (i.e., more people use an IBM mainframe than use an PC >[except, possibly, as a terminal to the mainframe]). Not sure I believe you about PC's, estimates are there are 15 to 30 million PC's out there and their use as terminals onto mainframes is usually bemoaned as "still waiting to happen". As far as Fortran and Cobol, again, how do you know this? Something I've used to measure this latter claim is to take the Sunday Jobs section of a major newspaper (I've used the Sunday Boston Globe) and make a simple tick count of jobs being offered in various areas. Last I did it (86-88) Unix/C was catching up rapidly on traditional areas (IBM Mainframe, Cobol, Fortran, BAL.) I know, that's only because Unix/C is a growth area and that's what this is really measuring. My guess is that's like saying Maseratis usually win races only because they're fast. I suppose you could nitpick the measure but it would be far more productive to suggest a better measure (the nice thing about this one is that anyone can do it in their living room in a few minutes.) IBM's mainframe predominance in the computing world is shrinking rapidly, that's why their revenues are in trouble. There was a time when they accounted for as much as 80% of *all* computer sales in the world, just a few years ago. I don't think they account for 50% anymore (partly due to growth in the industry around them.) And the current sales are heavily weighted towards a relatively few customers (fortune 100, US Govt), not that their money isn't green, but it's not as widespread an environment as it was say 10 years ago, particularly in relative terms. 50% is nothing to sneeze at, but if one is trying to do predictions the trends are pretty clear. Either IBM comes up with something brilliant to buoy their mainframe market (something I wouldn't discount as a possibility) or expect a rapid decline over the next 5 or so years as people realize they can "down-size" effectively ("down-sizing" is a term used in the MIS/DP market for replacing mainframe facilities with smaller machines, if you read that press you'd be shocked how many blue-serge suits are standing up and making testimonials about how they're decommissioning, or planning to in the near future, their mainframes and moving to a PC/LAN network with maybe a mini, typically AS/400 or Vax, at the hub.) There are areas where down-sizing doesn't cut it, but there are a *lot* of areas where it does, and that's skimming the cream out of the market. I suppose what people are talking about here is "down-sizing" in the super-computer market. It makes a lot of sense, most scientists I've worked with (I used to be in charge of most of the computers at BU and earlier, for a short while, ran the Harvard Chemistry computing facility, before that I was at the Harvard School of Public Health for several years) seem to prefer having political control over smaller facilities rather than go begging to centralized administrators. As under $100K systems approach 100MIPs and 30MFLOPS or so I don't see where the motivation to hassle with a system that's shared by hundreds of people will come from, except for those perhaps several dozen groups in the country that absolutely must be on a super-computer, even they'll do more and more prototyping and development on department-sized or personal facilities. Again, dwindling numbers. It's already happening. -- -Barry Shein Software Tool & Die, Purveyors to the Trade | bzs@world.std.com 1330 Beacon St, Brookline, MA 02146, (617) 739-0202 | {xylogics,uunet}world!bzs
mash@mips.COM (John Mashey) (11/26/89)
In article <3898@scolex.sco.COM> seanf@sco.COM (Sean Fagan) writes: >In article <39361@lll-winken.LLNL.GOV> brooks@maddog.llnl.gov (Eugene Brooks) writes: >>In article <3893@scolex.sco.COM> seanf@sco.COM (Sean Fagan) writes: >>>most "killer micros" is defficient because I can't do *real* DMA (it tends >>>to steal cycles from the CPU). (N.B.: some K.M.'s *do* have *real* DMA. >>>I'm waiting for them to come out with *real* I/O subsystems [using, say, a >>>68000 as a PP]. Then they will scream, even compared to a Cyber.) Just to correct a potential mis-impression: a) Since 1983 (or earlier, in a few cases, I think), anybody seriously building multi-user systems / servers from microprocessors has tended to build at least the high end of a product range with micros [68K, 186s, Z8000s, V-??, etc] as I/O processors. Some workstations [Sony News, for example] have 2 68Ks, one as an I/O processor. b) Although one may choose to use a "Killer Micro" in a workstation/PC/cheap server architecture, where there may be only one path to a memory bus with SIMMs, or similar design: 1) It usually has DMA. 2) it usually has a cache, and so I/O has some impact, but is hardly what people used to call cycle-stealing (where every I/O stopped the CPU almost cold). c) Any "Killer Micro" aimed at larger server/multi-user designs (as opposed to least-cost designs): has DMA usually has CPUs in at least some of the I/O boards, where appropriate sometimes has multiple paths to memory, i.e., a VME I/O bus and a private memory bus d) Many of the current high-performance I/O boards have 68020s, already, as in some of Interphase's products.
sbf10@uts.amdahl.com (Samuel Fuller) (11/28/89)
In article <1989Nov25.200320.21142@world.std.com> bzs@world.std.com (Barry Shein) writes: >Either IBM comes up with something brilliant to buoy their mainframe >market (something I wouldn't discount as a possibility) or expect a >rapid decline over the next 5 or so years as people realize they can >"down-size" effectively ("down-sizing" is a term used in the MIS/DP >market for replacing mainframe facilities with smaller machines, if >you read that press you'd be shocked how many blue-serge suits are >standing up and making testimonials about how they're decommissioning, >or planning to in the near future, their mainframes and moving to a >PC/LAN network with maybe a mini, typically AS/400 or Vax, at the >hub.) There are areas where down-sizing doesn't cut it, but there are >a *lot* of areas where it does, and that's skimming the cream out of >the market. > The opposite is also happening. Many large corporations are consolidating their data processing into larger centers. Shutting down several regional processing centers in favor of one large national center. Partitioning large mainframes has also become very arrtactive to some corporations. One big machine is logically partitioned into several smaller machines. Software and maintainance costs are lower for one large machine than for a half dozen small machines. -- --------------------------------------------------------------------------- Sam Fuller / Amdahl System Performance Architecture I speak for myself, from the brown hills of San Jose. UUCP: {ames,decwrl,uunet}!amdahl!sbf10 | USPS: 1250 E. Arques Ave (M/S 139) INTERNET: sbf10@amdahl.com | P.O. Box 3470 PHONE: (408) 746-8927 | Sunnyvale, CA 94088-3470 ---------------------------------------------------------------------------