ted@imsvax.UUCP (Ted Holden) (12/22/86)
Tim Kay of Caltech writes: >(Ted Holden) writes: >>The clones cost about half or a third what IBM's do and they (the >>clones) don't break. IBM must be wishing they'd never heard the words >>"PC" or "micro-computer" along about now. Without IBM's interference, >>micros would never have achieved the standardization which is now >>allowing them to challenge minis and mainframes. And IBM? They >>invented the PC/DOS game and now they can't even play their own game >>successfully and the game is threatening to destroy their big Fortune >>500 mainframe business. Kind of like letting the genie out of the >>bottle. >You've said several very interesting things here. First of all, >which is more reliable, IBM or compatibles? Business people often >think that, if you buy a machine from IBM, it must be more reliable >than a machine from some small company. And I have seen many a flakey >clone. However, I have also seen many lemons from IBM. Big Blue seems >to have very poor quality control regarding their PCs. They also offer >(by recent standards) an unreasonably short warranty and a very >expensive maintenance contract. I am beginning to think that IBM >equipment might cost more to keep running. The pieces for PCs are all pretty much generic now; anybody can fix them, and most do so for less than IBM charges. IBM owners have suffered major grief with CMI disks in recent times (check the April 29,86 issue of PC Magazine for details), and their system of subcontracting is not such as to allay fears that similar things won't be happening on other models in the future. >Next, I can't see how PCs are competing with minis and mainframes. An >80[23]86 at 8 or even 16Mhz still doesn't pack a fraction of the >computing power of a Vax 11/780. And, for the work I do, a Vax is >a small machine. A 3090/400 is roughly 50 times as powerful. If your name is John Rockafeller and you go out and purchase a 3090 as YOUR personal computer, then and only then does your argument make any sense. In real life, it never works like that. You will always be one of 300 people trying to use a mainframe at the same time, and the legitimate comparison is between an AT and whatever fraction of a mainframe's capabilities you are ever likely, in the real world, to be able to use. The comparison is between a mainframe and a system of micros. Four years ago there may have been some break-even point beyond which the multi-user machine was cheaper on a per user basis; now it isn't even close. A good 8mh XT compat with a hard disk can be had now for less than the cost of a terminal to a mainframe. Especially for applications which are screen I/O intensive, the idea of using multi-user computers no longer makes any sense. A VAX set up to serve 30 people doing mostly word processing will cost $150K - $200K, including terminals, wiring, and software. It'll be slower than hell, you'll have 30 people out looking for jobs all the time and, whenever it goes down, you'll have 30 people sitting on their thumbs for two or three whole days. $70K, intelligently spent, could have XT class machines on everybody's desk with legal copies of WordPerfect and good dot-matrix printers, two laser printers for everybody to share, and two or three back-up machines off to one side so that nobody ever sits on their thumbs when a machine goes down; you just swap one of the extras for the machine out being fixed. Development environments in which most users' time is spent in text editors present an entirely similar situation. I have never seen anyone smile while using a mainframe the way they do using ATs. There are many good reasons. Memory for multi-user machines has always been expensive and scarce; you no sooner try to scroll down one page in an editor or word processor on a mainframe than you have to swap information in from disk and you're ALWAYS 300'th in line to use that disk. Memory for PCs is dirt cheap and DOS programs such as WordPerfect and SuperCalc reflect that and USE it as if it were plentiful. WordPerfect can scroll through a 50 or 100 page document in seconds; I don't know of any mainframe product which can. Turbo Pascal can compile 2000 line programs in 10 seconds or so on a 6mh AT; again, I don't know of any mainframe compiler which can do this. The superior compute power claimed for mainframes has a way of seeming very little in evidence in the real world, in which most of the speed and elegance which users actually observe derives from intelligent programming rather than raw power. And the best programming being done these days is for the mass market machines where the biggest payoff is, not for mainframes. Likewise, hardware breakthroughs are now hitting the DOS market first and only then possibly filtering down to the mainframe market. This includes all kinds of things from superior disk technology to optical scanners, projection devices etc. As I see it, the day of the expensive computer is about over. It is only for super-computer applications such as weather forecasting and really big database applications that they could be justified at all any more, and the small machines will be capable of those activities in another couple of years.
news@cit-vax.Caltech.Edu (Usenet netnews) (12/22/86)
Organization : California Institute of Technology Keywords: mainframes From: tim@tomcat.Caltech.Edu (Tim Kay) Path: tomcat!tim [ Before we continue, I realize that the subject here is slightly tangent to the recent typical use of this newsgroup. If you feel that it is inappropriate here, let us know. I believe that it is a reasonable topic in the grander scope of this newsgroup, as it addresses the future of PC's versus mainframes. Tim ] In article <653@imsvax.UUCP> ted@imsvax.UUCP (Ted Holden) writes: > >Tim Kay of Caltech writes: > >>(Ted Holden) writes: >>>The clones cost about half or a third what IBM's do and they (the >>>clones) don't break. IBM must be wishing they'd never heard the words >>>"PC" or "micro-computer" along about now. Without IBM's interference, > >>Next, I can't see how PCs are competing with minis and mainframes. An >>80[23]86 at 8 or even 16Mhz still doesn't pack a fraction of the >>computing power of a Vax 11/780. And, for the work I do, a Vax is >>a small machine. A 3090/400 is roughly 50 times as powerful. > >If your name is John Rockafeller and you go out and purchase a 3090 as >YOUR personal computer, then and only then does your argument make any > >The comparison is between a mainframe and a system of micros. Four years ago >there may have been some break-even point beyond which the multi-user machine >was cheaper on a per user basis; now it isn't even close. A good 8mh XT compat > >As I see it, the day of the expensive computer is about over. It is only for >super-computer applications such as weather forecasting and really big database >applications that they could be justified at all any more, and the small >machines will be capable of those activities in another couple of years. I agree completely with you that, IF you want to do word processing or spreadsheets or program development, then micros (specifically IBM PC's and compatibles) are MUCH nicer than mainframes. They may not always be more practical, though. 1. typically response time is usually far better on PC's 2. software on PC's is far better than on mainframes 3. PC's hardware is more "efficient" -- per mips or flops a PC costs less than a mainframe Your main premise is that the majority of computer usage in this country has to do with personal computer type of usage. This is simply not the case. As PC's became available, people haven't moved existing work off of mainframes onto PC's. They have created new work to be done, or they have automated tasks that used to be done by hand. The work that the mainframes used to do is still being done on mainframes. Nobody used a spreadsheet program until there were (Apple) personal computers. I would guess that 99% of the computer applications that most people are aware of account for a tiny fraction of the total computer usage in the world (or even this country). IBM's bread and butter is derived from selling computers for uses which most people wouldn't even begin to imagine. As an example, "Communications of the ACM" last year or the year before did a case study of the TWA reservation system. The article said (I am recalling from memory) that the computer system handles about _seven_million_transactions_ _each_day_. This is a fundamentally different way of using computers than you were considering above. There is no possible way that personal computers will ever be able to do this sort of thing. (Why would you want them to?) Databases of this type are very common among large companies. I would guess that IBM makes as much profit selling a half dozen 3090's as they do selling an entire year's worth of PC's. (And they sell many hundreds of 3090's.) I will look up actual numbers if you are interested. To summarize, you are right that, for the limited applications that you selected, PC's are better. However, they represent only a very tiny piece of the total computer-usage pie. For other reasons, I do believe that IBM has seen its best days. This is only indirectly due to the advent of personal computers. In general, the science of software development has developed over the last couple of decades so that software manufacturer's are finally capable of writing software that is relatively independent of the underlying computer architecture. This means that other hardware vendors can once again innovate and still be able to sell hardware to companies that are currently using IBM machines. Therefore, I think that IBM's market share will erode very slowly over the next decade or two until IBM takes the status of Univac or Burroughs. This is unless they do something drastic, which I don't currently see them doing. Think about it: what was the last new innovation that IBM marketed? When was the last time they went out on a limb? Since they introduced the S/360 architecture, they haven't done anything first. They introduced the RT as a reaction to Sun and Apollo. They added the vector feature to the 3090 as a reaction to Cray. They added the PC as a reaction to Apple (and in that one case, their reputation helped them. They also got lucky). Everything else they sell simply services their customers that are locked in because of compatibility concerns. For this reason, and not because of PC's, is IBM's stock becoming shaky. I look forward to reading other opinions. Timothy L. Kay tim@csvax.caltech.edu Department of Computer Science Caltech, 256-80 Pasadena, CA 91125
mat@amdahl.UUCP (Mike Taylor) (12/22/86)
In article <653@imsvax.UUCP>, ted@imsvax.UUCP (Ted Holden) writes: > >Next, I can't see how PCs are competing with minis and mainframes. I can't either. There is no evidence that PCs have cut into or eroded mainframe capacity demand in any way, and some evidence that they have increased it. > > The comparison is between a mainframe and a system of micros. > I have never seen anyone smile while using a mainframe the way > they do using ATs. Maybe the problem here is comparing a 10-year-old VAX-11/780 with a modern micro, which isn't fair. I've certainly seen people smile when a several thousand line compile or build is done in seconds. Our current top-end product is about 60 times the capacity of a VAX-11/780 in a Unix (tm AT&T) environment. If your productivity is in any way related to turnaround or response on non-trivial computing tasks, then it's a better deal. In addition, many large tasks and databases simply aren't feasible on micros or systems of micros. The majority of mainframe computing is not, and never has been, small program development, spreadsheets, or document preparation. These are tasks for which PCs may indeed be well-suited. > WordPerfect can scroll through a 50 or 100 page document in seconds; > I don't know of any mainframe product which can. > > Turbo Pascal can compile 2000 line programs in 10 seconds or so on a 6mh AT; > again, I don't know of any mainframe compiler which can do this. > And the best programming being done these days is for the mass market machines > where the biggest payoff is, not for mainframes. > Clearly, your knowledge of mainframes is limited. However, you raise an interesting point. Actually, the biggest software payoff by far is in mainframes - the problem is that IBM dominates the market. It makes at least $2 billion a year from mainframe software - take that, Lotus. IBMs market clout has made it hard for others, specifically by use of a proprietary, closed operating system (MVS). But you are right - there is more clever programming by far being done for micros than for mainframes. But that's an opportunity! > Likewise, hardware breakthroughs are now hitting the DOS market first and only > then possibly filtering down to the mainframe market. You couldn't be more wrong on this one. Virtually all the ideas you see in micros were originally developed for mainframes - often many years ago. Certainly they've been repackaged for the micro - the value of a 1.2 GB hard disk on a PC is questionable - but you saw them here first. I would say that the only major technology that is more or less exclusively due to the micro is the bit-mapped display idea. Gee, pretty soon PCs will even have virtual memory and memory protection! -- Mike Taylor ...!{ihnp4,hplabs,amd,sun}!amdahl!mat [ This may not reflect my opinion, let alone anyone else's. ]
sam@lanl.ARPA (Sam A Matthews) (12/23/86)
[] To say who needs a mainframe when you can have a micro is like saying who needs a hammer now that we have a screw driver. :) The truth of the matter is both are tools and both do thier jobs very well. I have a PC on my desk that I use quite effectively. I also have a comm line that puts at my immediate disposal 17 vaxen minis and remote access that can get me to many of the 200+ vaxen (!) here at the lab. There are also 2 or 3 cray X-MP/48s, 1 cray X-MP/24, 4 cray-1s, 2 CDC 7600s, 5 CDC cybers, we also have 4 IBM mainframes (2-4241, 3083, and a 4381) just for a front end to mass storage! I would like to see someone try to get 1000+ users to access one database using a PC network or manage a multi-gigabyte database, or 2 or 3. Come on... It will not happen in our lifetimes. Technology is not just improving in the PC world either. Mainframe technology is also improving, advancing and growing in leaps and bounds. By the time we all have crays on our desks, I can just imagine what the crays will be doing! :) I guess the point of this is that tools are here for us to use, we should use the one that best does the job. I wouldn't use the cray for a spreadsheet any more than I would do fluid dynamics simulations on my PC. Sam Matthews /\|/\ "We put a star sam@lanl.ARPA --> * <-- in a box." (ihnp4 or cmcl2)!lanl!sam \/|\/
ben@catnip.UUCP (Bennett Broder) (12/24/86)
In article <653@imsvax.UUCP> ted@imsvax.UUCP (Ted Holden) writes: >The comparison is between a mainframe and a system of micros. Four years ago >there may have been some break-even point beyond which the multi-user machine >was cheaper on a per user basis; now it isn't even close. A good 8mh XT compat >with a hard disk can be had now for less than the cost of a terminal to a >mainframe. Especially for applications which are screen I/O intensive, the >idea of using multi-user computers no longer makes any sense. A VAX set up to >serve 30 people doing mostly word processing will cost $150K - $200K, including >terminals, wiring, and software. It'll be slower than hell, you'll have 30 >people out looking for jobs all the time and, whenever it goes down, you'll >have 30 people sitting on their thumbs for two or three whole days. > >$70K, intelligently spent, could have XT class machines on everybody's desk >with legal copies of WordPerfect and good dot-matrix printers, two laser >printers for everybody to share, and two or three back-up machines off to one >side so that nobody ever sits on their thumbs when a machine goes down; you >just swap one of the extras for the machine out being fixed. Development >environments in which most users' time is spent in text editors present an >entirely similar situation. You are leaving out one of the most important features of a multi-user system, sharing of data. What happens when one of your 30 people wants to do a mail-merge of data stored on another machine? What happens when the database on machine one says I live at 33 maple street, and on machine two says I live at 12 vine street? You have on your hands a disaster. The only way PCs make sense for a big company is if they are networked with a high speed (ethernet or equivilent) local area network. If you add the cost of networking hardware and software into the picture, the multiuser machines starts looking a lot cheaper. >I have never seen anyone smile while using a mainframe the way they do using >ATs. There are many good reasons. Memory for multi-user machines has always >been expensive and scarce; you no sooner try to scroll down one page in an >editor or word processor on a mainframe than you have to swap information in >from disk and you're ALWAYS 300'th in line to use that disk. Memory for PCs is >dirt cheap and DOS programs such as WordPerfect and SuperCalc reflect that and >USE it as if it were plentiful. WordPerfect can scroll through a 50 or 100 >page document in seconds; I don't know of any mainframe product which can. I have a hunch that your smiling AT users are not people who have to write big programs. Memory for the IBM-AT may be cheap, but it a pain in the **** to access. To someone who learned to program in 'C' on a vax, the AT is a 64k machine. Doubt what I say? I have an AT with several megs of extended memory. Send me a working pathalias. >As I see it, the day of the expensive computer is about over. It is only for >super-computer applications such as weather forecasting and really big database >applications that they could be justified at all any more, and the small >machines will be capable of those activities in another couple of years. You may be right. When I see the power of the 68030 and the 80386, I am astounded. I suspect that in a few years, you will have all the compute power you need, right in you own machine. But I think that the answer for today is blend of machines. A combination of PCs, multiusers PCs, minis and mainframes all linked together with a high speed network. That way, the micros can do what they do best (interactive stuff like spreadsheets) while the minis and mainframes what they do best (number crunching and storing gigabytes of data to be shared by all users). -- Ben Broder {ihnp4,decvax} !hjuxa!catnip!ben {houxm,clyde}/
sampson@smu.UUCP (12/29/86)
It appears that the new 32-bit microcomputers will have the CPU power of many smaller mainframes. The weakest link will be I/O. The disk access is much too slow for any moderate number of users. There is no doubt about the impact of PC's on the mainframe and especially minicomputer market. But, as long as there are mundane jobs that must be done in quanity, such as gas bills and payroll, there will always be mainframe shops around. In terms of micro users sharing data, the best method of networking pc's is too hook them all to a mainframe with a large secondary storage capacity. I wouldn't trade my micro even for a 3090, if I had to do ALL my work on the 3090, but there is a place for both. Not only will there always be mainframes, there will always be COBOL! joel sampson southern methodist university convex!smu!sampson
ken@argus.UUCP (Kenneth Ng) (12/29/86)
In article <10781@lanl.ARPA>, sam@lanl.ARPA (Sam A Matthews) writes: > [] > I would like to see someone try to get 1000+ users to access one database > using a PC network or manage a multi-gigabyte database, or 2 or 3. > Come on... It will not happen in our lifetimes. That's the same arguement I have here at school. People claim the vaxes are faster than the mainframe here on campus, but they fail to realize that the vaxes work with a load of 10, while the mainframe has a load of around 100. > Sam Matthews /\|/\ "We put a star > sam@lanl.ARPA --> * <-- in a box." > (ihnp4 or cmcl2)!lanl!sam \/|\/ -- Kenneth Ng: Post office: NJIT - CCCC, Newark New Jersey 07102 uucp !ihnp4!allegra!bellcore!argus!ken *** WARNING: NOT ken@bellcore.uucp *** !psuvax1!cmcl2!ciap!andromeda!argus!ken bitnet(prefered) ken@orion.bitnet Gillian: "Are you sure you won't change your mind?" Spock: "What's wrong with the one I have?"
rotheroe@convexs.UUCP (12/30/86)
> It appears that the new 32-bit microcomputers will have the > CPU power of many smaller mainframes. The weakest link will be > I/O. The disk access is much too slow for any moderate number of > users. Although minisuper, main supermini, and minis will always be with us (and at any given time the current models of them will be much more powerful than the current micro's), one has to wonder how something like a 68030 running at (guessing) 25MHz with (say) 128MByte memory would compare. If almost everything can fit in memory at a time, I/O speed is less of a worry. Such a micro could easily serve a dozen (or more) users, each running a (relatively) large job, and almost never have to go to disk. We will have to wait and see what happens, and until that time I'll stick to minisupercomputers. Dave Rotheroe {allegra, ihnp4, uiucdcs, ctvax}!convex!rotheroe "Good afternoon, gentlemen. I am a HAL 9000 computer. I became operational at the Hal plant in Urbana, Illinois, on the twelfth of January, 1992." 2001 & 2010 (book only for 2010)
james@bigtex.uucp (James Van Artsdalen) (01/02/87)
In article <10781@lanl.ARPA>, sam@lanl.ARPA (Sam A Matthews) writes: > I would like to see someone try to get 1000+ users to access one database > using a PC network or manage a multi-gigabyte database, or 2 or 3. > Come on... It will not happen in our lifetimes. Actually, a company called Products Diversified in Houston TX has been selling 68000-based micros for a while that can handle 8 gigabyte databases pretty well. Lew Williams founded the company many years back as a result of problems he had when he was in the land-title business: people couldn't sell him reasonably priced hardware for large database applications. It's been two or three years since I talked to Lew, and in any case his son-in-law manages the company, but the point is that they have been doing multi-gigabyte databases on 68000s, not even 68020s, for years (used a variant on the LSI-11 before the 68000). I believe they use a specialized black box that does the real work, and that the box costs in the neighborhood of $50,000 to $100,000. The thing has been networked a fair amount, but to get more info call PDI and ask about it. 1000 users maybe not, but an 8 gigabyte database on a micro is old news, not "will not happen". -- James R. Van Artsdalen ...!ut-sally!utastro!bigtex!james "Live Free or Die" Voice: (512)-323-2675 Modem: (512)-323-2773 5300B McCandless, Austin TX 78756
curt@charming.uucp (Curt Mayer) (01/03/87)
In article <653@imsvax.UUCP> ted@imsvax.UUCP (Ted Holden) bs's: >You will always be one of 300 people trying to use a mainframe at the >same time, and the legitimate comparison is between an AT and whatever >fraction of a mainframe's capabilities you are ever likely, in the real >world, to be able to use. Give me a break. I have never seen a CDC cyber with more than 05% load. This is in a university environment with 700+ users, no less. >As I see it, the day of the expensive computer is about over. It is only for >super-computer applications such as weather forecasting and really big database >applications that they could be justified at all any more, and the small >machines will be capable of those activities in another couple of years. have you ever seen a 5000 line program compile in less time than it takes for the print head to return on a DecWriter? Guess not. give mister J-Random Spreadsheeter his AT, I'll keep the real machines for real work. curt
rodney@gitpyr.gatech.EDU (Rodney Ricks) (01/03/87)
In article <196@unisoft.UUCP> curt@charming.UUCP (Curt Mayer) writes: >Give me a break. I have never seen a CDC cyber with more than 05% load. >This is in a university environment with 700+ users, no less. > What? Are you saying that you are using a Cyber system that can handle 700+ users at a time? My experience with a Cyber is with a machine that dies a horrible, crawling death when 125 users get on it. With this number of users, the machine would slow down to the point of compiling programs at the speed of a standard IBM PC,... with floppies! That Cyber was replaced with a newer Cyber system about a year ago. Since I haven't done much work on it lately (I avoid it. Almost everybody avoids it!), I don't know how much faster it is. By the way, something just came to mind about why your Cyber might be so much faster than our old Cyber. I think our Cyber had (only) two CPU's in it. How many does yours have? It does seem kind of unfair, comparing multi-CPU mainframes with single-CPU micros and saying "Look at how much faster the mainframes are!". Are there any other students out there who have experienced just how slow a Cyber can be (come on Georgia Tech students, speak up!)? > > curt Disclaimer: The above information is supplied without warranty, either expressed or implied, about its merchantability, fitness, or reliability for any purpose. To put it concisely, B E L I E V E I T O R N O T ! ! ! Rodney Ricks UUCP: ...!{akgua,allegra,amd,hplabs,ihnp4,seismo,ut-ngp}!gatech!gitpyr!rodney or : !gatech!gt-oscar!rodney Mail: 4265 Hidden Valley Dr. College Park, Ga. 30349
robert@gitpyr.gatech.EDU (Robert Viduya) (01/03/87)
>curt@charming.UUCP (Curt Mayer) (curt@charming.UUCP, <196@unisoft.UUCP>): > Give me a break. I have never seen a CDC cyber with more than 05% load. > This is in a university environment with 700+ users, no less. Is this 700+ simultaneous online users or just 700+ users in the validation file? Our main computing engines, two CDC Cyber 855s and one CDC Cyber 990, can only handle between a 100 and 200 users on simultaneously (per machine), although the validation file has thousands of users in it. robert -- Robert Viduya robert@pyr.ocs.gatech.edu Office of Computing Services (404) 894-4660 Georgia Institute of Technology Atlanta, Georgia 30332
news@cit-vax.Caltech.Edu (Usenet netnews) (01/04/87)
Organization : California Institute of Technology Keywords: From: tim@tomcat.Caltech.Edu (Tim Kay) Path: tomcat!tim In article <2839@gitpyr.gatech.EDU> rodney@gitpyr.UUCP (Rodney Ricks) writes: >In article <196@unisoft.UUCP> curt@charming.UUCP (Curt Mayer) writes: > How many does yours have? It does seem kind of unfair, comparing >multi-CPU mainframes with single-CPU micros and saying "Look at how much faster >the mainframes are!". That is exactly the point! There are people submitting to this newsgroup articles that claim that those same PCs (or their future models) will replace mainframes like the Cyber. Timothy L. Kay tim@csvax.caltech.edu Department of Computer Science Caltech, 256-80 Pasadena, CA 91125
bobmon@iuvax.UUCP (Robert Montante) (01/04/87)
> robert@gitpyr.UUCP (Robert Viduya): >>curt@charming.UUCP (Curt Mayer): >> Give me a break. I have never seen a CDC cyber with more than 05% load. >> This is in a university environment with 700+ users, no less. > >Is this 700+ simultaneous online users or just 700+ users in the validation >file? Our main computing engines, two CDC Cyber 855s and one CDC Cyber 990, >can only handle between a 100 and 200 users on simultaneously (per machine), >although the validation file has thousands of users in it. A couple of years ago I worked on an IBM 3090 (probably been upgraded again by now) that regularly had 400+ users on it. By regularly I mean that I got in at 7:30am to get to work because around 8:15am, when the Data Processing and Stat. Support people had gotten their coffee, the load would hit the 400+ number -- five days a week. I think the installation had three 3090's, only two of which were dedicated to interactive processing. Incidentally, I would guess that 60% of the terminals on that system were PC's. In my experience, the 400-user limit was defined by the system response time. At that number, response was poor enough that additional people would find something else to do. Another point: I would download whatever I could to the PC, thereby improving response time for me quite a bit. There was still quite a bit of work that couldn't be downloaded, either because the software was mainframe-resident only, or the data were mainframe-resident, or I was doing something that explicitly involved the networking features of the system. In short, my experience was that, in large environments, an explosive growth in the use of personal computers makes life marginally more bearable on the mainframes. There is still demand for all the mainframe capacity available, but more people struggle with the poor response times (instead of giving up entirely) because many of the smaller jobs move to the local machines. *-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-*-=-* Datclaimer: "[Usual disclaimer: I have no opinion, therefore I don't exist .]" Disclaimer: I opine, therefore I am. My employer, however, is a figment. RAMontante Computer Science "Have you hugged ME today?" Indiana University
johnl@ima.UUCP (John R. Levine) (01/05/87)
In article <1416@cit-vax.Caltech.Edu> tim@tomcat.UUCP (Tim Kay) writes: >That is exactly the point! There are people submitting to this newsgroup >articles that claim that those same PCs (or their future models) will >replace mainframes like the Cyber. And for some purposes I'm sure they eventually will. For example, the real estate transfer records in Houston which comprise a multi-gigabyte data base are maintained on a Britton-Lee data base machine with queries being made from minimal PCs at 300 baud. The performance is apparently wonderful, since the size of the requests and responses tend to be small. The largest procesor involved is a Z8000. This is a job that a few years ago would clearly have needed a mainframe, but doesn't any more. It's true -- PC peripherals tend to be junk, and the software is worse. (I can vouch for the software, I write it for a living.) But it is all evolving much faster than the mini or mainframe equivalent, so don't rule anything out. I put a Fujitsu Eagle on a PC over a year ago (and the idiot software made me partition it as logical drives D: to O:) and I'm sure I can do better now. Real operating systems for PCs are gasping their way to reality now (and before you sneer, remember the early days of VMS or, perish forbid, OS/360) and will likely take off as users buy 386 boxes that can support reasonable multitasking and protection. Compare a current 386 box to the micro of 1977, and then compare a 3090 mainframe to the mainframe of 1977. Which one has changed more? Which one has gained more power? Which one would you bet on for 1997? Apocalyptically, -- John R. Levine, Javelin Software Corp., Cambridge MA +1 617 494 1400 { ihnp4 | decvax | cbosgd | harvard | yale }!ima!johnl, Levine@YALE.something Where is Richard Nixon now that we need him?
kds@mipos3.UUCP (Ken Shoemaker ~) (01/07/87)
actually, I'd really like to use our mainframe as a fileserver to my
workstation. Does anyone out there know if anyone does this yet?
--
The above views are personal.
The primary reason innumeracy is so pernicious is the ease with which numbers
are invoked to bludgeon the innumerate into dumb acquiescence.
- John Allen Paulos
Ken Shoemaker, Microprocessor Design, Intel Corp., Santa Clara, California
uucp: ...{hplabs|decwrl|amdcad|qantel|pur-ee|scgvaxd|oliveb}!intelca!mipos3!kds
csnet/arpanet: kds@mipos3.intel.commadd@bucsb.bu.edu.UUCP (Jim "Jack" Frost) (01/13/87)
In article <53900006@smu> sampson@smu writes: > It appears that the new 32-bit microcomputers will have the >CPU power of many smaller mainframes. The weakest link will be >I/O. The disk access is much too slow for any moderate number of >users. Wrong! Fancy programming can take care of this. We operate a multiuser multiprocessor PC system. Now, with 5 terminals running I/O intensive stuff all at once, there is almost no slowdown. Why? Disk caching. We have a 1.5 megabyte cache. It really makes a difference. It's write- through, so if everyone is writing at the same time, things slow down. But we bought a $1200 micropolis drive (25ms ave access advertised) which is plenty fast enough for a 5 user system with the cache. Some data compression techniques I've seen can also speed up the apparent I/O speed by compressing the actual amound of data being read or written. > There is no doubt about the impact of PC's on the mainframe >and especially minicomputer market. But, as long as there are mundane >jobs that must be done in quanity, such as gas bills and payroll, there >will always be mainframe shops around. In terms of micro users sharing >data, the best method of networking pc's is too hook them all to a >mainframe with a large secondary storage capacity. Basically I agree with that. There's no way a PC can match an IBM 3090 (yet) in pure processing power. And there's better on the way. But with some of the multi-CPU architectures they have (like the one we use) you can get the megaMIPS that the 3090 supplies without the megacost. This could wipe out IBM's System/36 market. I did a cost projection for a multi-CPU system with the processing power of the 3090 and came out with a system in the System/36 price range, including software and terminals. Pretty sad state of affairs for IBM. Still out of the PC price range, though. > I wouldn't trade my micro even for a 3090, if I had to do >ALL my work on the 3090, but there is a place for both. Me either, but luckily I get to play with both. Kinda like having your cake and eating it too :-) >Not only >will there always be mainframes, there will always be COBOL! Yea, but will anyone use it? RPG II is still around, but who do you know that uses it? I'm one of the few people here that had even HEARD of it, and the only one I've met here who can write serious stuff in it. The language is around, but the programmers aren't. It's just a matter of time before it costs too much to keep the language in stock (and maintained). >joel sampson >southern methodist university >convex!smu!sampson %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% - Jim Frost * The Madd Hacker - UUCP: ..!harvard!bu-cs!bucsb!madd | ARPANET: madd@bucsb.bu.edu CSNET: madd%bucsb@bu-cs | BITNET: cscc71c@bostonu -------------------------------+---+------------------------------------ "Oh beer, oh beer." -- Me | [=(BEER) <- Bud the Beer (cheers!)