nash@ucselx.sdsu.edu (Ron Nash) (02/24/90)
I am throwing this to the collective net wisdom (:-). I am asking for suggestions and any experiences good or bad to guide us in our search. The main uses of the computer will be teaching Unix and programming, and research. We have Macsyma and Lisp users and neural-network research that can be quite a drain. We are looking for a Unix engine that will support at least 100 concurrent users. It would be nice if it could be expanded to support more users if needed. The Unix should be BSD or be BSD compatible. A major concern is compatability of BSD software. Please post or email. If there I get or see requests for a summary, I will post a summary of the responses. If you want any more info from me, please let me know. Thanks in advance! -- Ron Nash San Diego State University Internet: nash@ucselx.sdsu.edu UUCP: ucsd!sdsu!ucselx!nash
neil@uninet.cpd.com (Neil Gorsuch) (02/25/90)
In article <long-id@ucselx.sdsu.edu> nash@ucselx.sdsu.edu (Ron Nash) writes: >I am throwing this to the collective net wisdom (:-). I am asking for >suggestions and any experiences good or bad to guide us in our search. >The main uses of the computer will be teaching Unix and programming, >and research. We have Macsyma and Lisp users and neural-network >research that can be quite a drain. >We are looking for a Unix engine that will support at least 100 >concurrent users. It would be nice if it could be expanded to support >more users if needed. The Unix should be BSD or be BSD compatible. A >major concern is compatability of BSD software. Here's one way to do it that will definitely save you big bucks. Instead of buying a single BIG machine for a BIG price, look to where the best deal is in (MIPS+Mbytes/$), which is currently desktop workstations, and buy some of those, TO BE USED AS MULTI-USER MACHINES. For instance, how about using 10 Sparcstations (or substitute your own favorite brand) as multi-user machines. Actually, I would only put in 5 and put 20 users per machine, but you might want to be more conservative and put in 10 with 10 users each. If you make each one a "dataless" client of an appropriate server, you will have a VERY peppy 10 (20) user machine. There's a nifty way to add a bunch of serial ports to desktop workstations without using the ethernet, so that shouldn't be a problem. Each Sparcstation is 12.5 MIPS or so, can have up to 64 Mbytes of memory, can have Gigabytes of 16 mS disk storage, and is BSD ported/compatible, so they should handle your needs. And with recent/upcoming announcements, you will probably be putting in 27 MIPS machines instead of 12.5 MIPS machines if you wait a few months 8-). Personally, I would even skip the server and sprinkle the user's files amoungst the workstations, with various NFS cross-mounts, while making sure to put as many user's files on the workstations that they mostly log onto as is possible. You should be able to do that with a lot of the students and most of the professors. You can be very creative in how you cross-mount user's home directories, while still backing everyone up each night over the network with a few Exabyte tapes. The entire "system" will be much more reliable, since you will have a number of machines so that one machine going down will not cripple all of your users. With the money that you save on a maintanence contract for a big machine, you could even have spare workstations lying around, and have users files duplicated on more than one disk, with automatic nightly duplication, so that in the event of a disk or workstations crash you could just switch around your NFS mountings for fast "repair". The system is very easily expandible, just add more workstations. In the case of the Sparcstations, you can even have 2 ethernet interfaces in each for ease of networking configuration. When researchers clamor for a workstation instead of a terminal, you can accomodate them without bogging down your primary "system" ethernet backbone by putting the private workstations on a second ethernet interface in the workstation that was their "login" computer. -- Neil Gorsuch INTERNET: neil@uninet.cpd.com UUCP: uunet!zardoz!neil MAIL: 1209 E. Warner, Santa Ana, CA, USA, 92705 PHONE: +1 714 546 1100 Uninet, a division of Custom Product Design, Inc. FAX: +1 714 546 3726 AKA: root, security-request, uuasc-request, postmaster, usenet, news
pjg@autarch.acsu.buffalo.edu (Paul Graham) (02/27/90)
In article <1990Feb25.075337.22513@zardoz.cpd.com>, neil@uninet (Neil Gorsuch) writes: |In article <long-id@ucselx.sdsu.edu> nash@ucselx.sdsu.edu (Ron Nash) writes: |>We are looking for a Unix engine that will support at least 100 |>concurrent users. It would be nice if it could be expanded to support |>more users if needed. The Unix should be BSD or be BSD compatible. A |>major concern is compatability of BSD software. | |Here's one way to do it that will definitely save you big bucks. |Instead of buying a single BIG machine for a BIG price, look to where |the best deal is in (MIPS+Mbytes/$), which is currently desktop |workstations, and buy some of those, TO BE USED AS MULTI-USER |MACHINES. [ and proceeds to describe a cluster of workstations acting as a timesharer] last semester we went looking for a timeshare machine. when we settled on a SPARC compatible system it was suggested that we consider purchasing a roomful of 4/60s rather than spend the money on a 4/490 and a solbourne 5/802. my answer then and even more so now is bzzzzt, wrong and no prize for you ;-). i wouldn't think for a second of putting 10 users on anything like a 4/60 let alone 20 (my workstation is a 4/60, with 12MB and the fast disk). i also think such a cluster would be nightmare to manage without having the users files on a server. now i do think the idea of multiple cpus is a good one and my experience with the 2 cpus in our solbourne has been very good (the machine can do almost 6000 context switches per second. this only the first semester we've encouraged a few people to run classes on it but it's been quite snappy with 60-70 users most of whom are running emacs and many of which are running ibcl) some hardware problems not withstanding. sunOS is, of course, BSD similar. you can also get macsyma et.al. for sparc machines. other reasonable sounding solutions would be an encore (too pricey for us) or perhaps a multi-cpu 5000 from dec, assuming ultrix still looks more like bsd than sysV. these alternatives are dependent on software of course. i've become a big fan of shared-memory multi-cpu boxes. i think they're clearly the wave of future. why even sun will have one someday. one just needs to make sure that each cpu is big enough and that the system design doesn't make peripheral/memory access a bottleneck. impertinent details: our budget was $250k (we came pretty close) we bought a 4/490 and 5/802 each with 64MB and about 2G of disk. we don't use serial ports and we didn't have to pay for terminal servers or maintenance (someone else's budget). we spent about the same amount of money on each machine and yes the solbourne is (with two cpus) twice as fast as the 4/490. we don't like putting all our eggs in one basket. all prices were for last summer. oh yah, 64 pty's isn't nearly enough and you have to teach rlogin/telnet about the rest.
neil@uninet.cpd.com (Neil Gorsuch) (02/27/90)
In article <18375@eerie.acsu.Buffalo.EDU> pjg@autarch.acsu.buffalo.edu (Paul Graham) writes: >In article <1990Feb25.075337.22513@zardoz.cpd.com>, >neil@uninet (Neil Gorsuch) writes: >|In article <long-id@ucselx.sdsu.edu> nash@ucselx.sdsu.edu (Ron Nash) writes: >|>We are looking for a Unix engine that will support at least 100 >|>concurrent users. It would be nice if it could be expanded to support >|Instead of buying a single BIG machine for a BIG price, look to where >|the best deal is in (MIPS+Mbytes/$), which is currently desktop >|workstations, and buy some of those, TO BE USED AS MULTI-USER >|MACHINES. >[ and proceeds to describe a cluster of workstations acting as a timesharer] >last semester we went looking for a timeshare machine. when we >settled on a SPARC compatible system it was suggested that we consider >purchasing a roomful of 4/60s rather than spend the money on a 4/490 >and a solbourne 5/802. my answer then and even more so now is bzzzzt, >wrong and no prize for you ;-). i wouldn't think for a second of putting >10 users on anything like a 4/60 let alone 20 (my workstation is a >4/60, with 12MB and the fast disk). i also think such a cluster would >be nightmare to manage without having the users files on a server. Methinks you don't know how to configure a Sparcstation for maximum performance, or that you haven't tried one with > 16 Mbytes yet 8-). A Sparcstation with 32 Mbytes or more of memory and a couple of 16 mS disks is VERY snappy for 10 or so users (or use 11 mS disks if you really want to be impressed. And as for system administration, just designate one of the machines as your main machine, and use rdist or rcp or whatever for configuration files and local binaries. Since you essentially have clone machines except for user's files, it should be easier to manage than 2 different types of machines (some programs need to be re-compiled between the Solbourne and the Sun, and they didn't do exactly the same things for systems directory trees). Sprinkle the users's files appropriately, with NFS mounts everywhere in case a user moves around, and use Yellow Pages to cut down on confusion. Backups are easy with a shared Exabyte drive. Please note that I have a Solbourne and a Sparcstation in my office, and I do admit that I use the Solbourne as my main machine. But we're talking about the best bank for buck for a time sharing machine. As for a 4/490, it's basically a somewhat faster Sparcstation with an overpriced card cage. But for multi-user systems, I would much rather have 3 or 4 Sparcstations with 32 or even 64 Mbytes each than a 4/490 any day, and so would the users that you divvy up per $ spent, if they could compare both systems. The Solbournes do real well at MIPS/$, but their memory is proprietary with no third party availability that I know of, while the Sparcstations have dirt cheap memory available elsewhere. Of course at 27 MIPS/$7450 list for a Data General Aviion, they come in even better. I know of a University that has done some investigating on this subject and found that they can't touch the price/performance of desktop workstations as multi-user systems after they add up all the costs of disks and memory and serial interfaces, and have even gone so far as to REQUIRE all future multi-user systems campus-wide to be DECstations, at least until some new workstation models are out. What you have done is choose a Solbourne (a very good choice 8-) for part of your "system", but then gone and bought an overpriced card cage that is expensive to add memory and disks to for the second part. >i've become a big fan of shared-memory multi-cpu boxes. i think >they're clearly the wave of future. why even sun will have one >someday. I completely agree, but for $'s spent, you probably can't beat the desktops, and remember we're talking about a wave of 27 MIPS desktops coming out in the next quarter or so. Not to mention the Taiwan Sparc machines due out soon. >impertinent details: our budget was $250k (we came pretty close) we >bought a 4/490 and 5/802 each with 64MB and about 2G of disk. Perhaps the problem last summer was getting past 16 Mbytes in a Sparcstation cheaply, but consider these costs now, assuming that you have 70 MIPS in your "system": 6 Sparcstations with 8 Mbytes Sun memory each $54,000 6 X 16 Mbytes (4 Mbyte SIMMS) $24,000 (approx.) 6 X 4 Mbytes (1 Mbyte SIMMS) $2,500 6 X 1.3 Gbytes shoeboxes (2 660 Mbyte disks each) $36,000 1 Exabyte and 1 1/4" tape drive $4,500 misc costs (software, etc.) $5,000 70 MIPS, 168 Mbytes, 7.9 Gbytes $126,000 And if I went ahead and spent a hair over $250,000, I would end up with a "140 MIPS, 336 Mbytes, 15.8 Gbytes" system. And if I was really gutsy and used 10 Aviions instaed of Suns, for the same price, I can bump it up to about 270 MIPS, and none of your 70 users is going to complain about getting only 1/7th of a 27 MIPS machine 8-). The trick is in finding desktop workstations that can have a decent amount of cheap memory and disks and MIPS. Hopefully, a lot more desktop workstations will have multiple CPU boards, and they will come out better in the equations. If I had my Solbourne list prices handy, maybe a 2 processor series 5 would be better overall, but I doubt it with the price of their memory. >we don't use serial ports and we didn't have to pay for terminal servers >or maintenance (someone else's budget). we spent about the same I assume that someone else is providing ethernet based terminal servers. The other side of this equation is that someone else spent a fair amount for your serial ports, and that with 65 people banging away through the ethernet, you waste a lot of ethernet bandwidth and host CPU time with the TCP overhead. Whereas if you split the system into smaller chunks with a bunch of serial ports directly in each of the Sparcstations you can optimally place the serial lines, you will save a noticable amount of $'s, and the hosts will be saving a noticable amount of CPU time. Not to mention that the users won't have to log in through an ethernet server and THEN rlogin into one of the hosts. >we don't like putting all our eggs in one basket. But you put it in 2 baskets, where I am proposing putting it in 6 baskets, which is even more reliable. I'm not saying your choice is bad (I love Solbourne), but I am saying that if you look at things in a different light, you perhaps could have spent about half of what you spent. Don't let a Sparcstation with too little memory (12 Mbytes is barely past the Sunos/sparc performance knee) sour you on the possibilities that a uniquely configured Sparcstation can offer 8-). And don't get the idea that I am pushing Sparcstations, they just happen to be a very cost effective solution right now, plus I know that there will be very cheap double digit MIPS versions out from Taiwan soon. -- Neil Gorsuch INTERNET: neil@uninet.cpd.com UUCP: uunet!zardoz!neil MAIL: 1209 E. Warner, Santa Ana, CA, USA, 92705 PHONE: +1 714 546 1100 Uninet, a division of Custom Product Design, Inc. FAX: +1 714 546 3726 AKA: root, security-request, uuasc-request, postmaster, usenet, news
long@castor.csg.uiuc.edu (Junsheng Long) (03/03/90)
I would like to know the data segment layout of a C program on Unix (Sun OS). For the following example, I found some extra variables were inserted before the first global variables (marked as ? in the following figure). /* --- example.c ---*/ +----------+ int a = 0; | text | int b[32]; +----------+ <- etext | ? | main() +----------+ { | a | ... +----------+ <- edata } | b | +----------+ <- end | | Now my questions are: does anyone know what they are? Does they vary from program to program? Thank you very much. junsheng long
pjg@autarch.acsu.buffalo.edu (Paul Graham) (03/03/90)
In article <1990Feb27.071903.578@zardoz.cpd.com>, neil@uninet (Neil Gorsuch) writes: 'In article <18375@eerie.acsu.Buffalo.EDU> pjg@autarch.acsu.buffalo.edu (Paul Graham) writes: [someone is looking to support 100+ users on a box that runs a variety of of useful software. neil gorsuch suggests getting a bunch of workstations, i suggest getting a larger computer.] '>last semester we went looking for a timeshare machine. [actually it was last summer that we decided to get a sun 490 and solbourne (sun-4 compatible) series 800 with 2 cpus. i note that i don't think much of timesharing on a sun 4/60 (sparcstation 1).] '[neil responds] 'Methinks you don't know how to configure a Sparcstation for maximum 'performance, or that you haven't tried one with > 16 Mbytes yet 8-). our hardware decisions are somewhat constrained. we tend not to buy things from companies like uninet. not a value judgement just our policy. 'A Sparcstation with 32 Mbytes or more of memory and a couple of 16 mS 'disks is VERY snappy for 10 or so users hmmmm, mine has 12M and a wren4, which is a 16ms drive i believe, and it's just this side of not irritating for me. 10 of our emacsing/lisping users would be a bit much. [neil suggests cloning the machines with rdist. we have a lot of per cpu licensed software. a real killer unless you get very sweet deals.] '(some programs 'need to be re-compiled between the Solbourne and the Sun, and they 'didn't do exactly the same things for systems directory trees). that's certainly not our experience (except for allegro and ibuki lisp on the 490). 'Sprinkle the users's files appropriately, with NFS mounts everywhere 'in case a user moves around, and use Yellow Pages to cut down on 'confusion. Backups are easy with a shared Exabyte drive. and raise the probabilty that any one dead box will hang the entire group. sun does claim that they've put some work into this for 4.1 though. (i do believe i hate nfs, although not for the same reasons as henry spencer) '. . . we're 'talking about the best bank for buck for a time sharing machine. my point exactly. 'As for a 4/490, it's basically a somewhat faster Sparcstation with an 'overpriced card cage. But for multi-user systems, I would much rather 'have 3 or 4 Sparcstations with 32 or even 64 Mbytes each than a 4/490 'any day, and so would the users that you divvy up per $ spent, if they 'could compare both systems. i assume that this is simple hyperbole. you really can't compare a machine with a large fast cache, 6M/sec disks, adequate support for context switching and the ability to support a lot of fast memory to a desktop. 'The Solbournes do real well at MIPS/$, 'but their memory is proprietary with no third party availability that 'I know of, while the Sparcstations have dirt cheap memory available 'elsewhere. most high performance memory systems are somewhat proprietary. memory per box is only one cost. [neil mentions a university that has concluded that workstations are best price/performance when you consider disk/ports/memory per $] 'What you have done is choose a Solbourne (a very good 'choice 8-) for part of your "system", but then gone and bought an 'overpriced card cage that is expensive to add memory and disks to for 'the second part. both are about the same price. we got the solbourne for price/performance, we got the sun because it was the best they had and sun will be in business in 2 years while solbourne (and uninet) may not. the 490 is ``extra''. [i mumble about how great multi-cpu boxes are] 'I completely agree, but for $'s spent, you probably can't beat the 'desktops, and remember we're talking about a wave of 27 MIPS desktops 'coming out in the next quarter or so. Not to mention the Taiwan Sparc 'machines due out soon. right, but i can use the same cpu in my multi box and only buy the memory and disk once. or i can get an even better cpu, much better disks and more memory available in *one* place and get to reduce my per box costs to 1 or 2. not to mention the ecl 100 MIPS chips due out in your nearby compute server soon. [we spent ~$250 on our two boxes] '. . . consider these costs, assuming that you have 70 MIPS in your "system": '6 Sparcstations with 8 Mbytes Sun memory each $54,000 '6 X 16 Mbytes (4 Mbyte SIMMS) $24,000 (approx.) '6 X 4 Mbytes (1 Mbyte SIMMS) $2,500 '6 X 1.3 Gbytes shoeboxes (2 660 Mbyte disks each) $36,000 '1 Exabyte and 1 1/4" tape drive $4,500 'misc costs (software, etc.) $5,000 '70 MIPS, 168 Mbytes, 7.9 Gbytes $126,000 i'm afraid our software costs would be a bit more than that *per* machine. we have (or will have shortly) modula, ada, two kinds of lisp, saber-c, macsyma, mathematica, matlab, lat and various other things that we pay for. and we could have bought boxes full of those scsi disks if we wished and saved a fair bit of money. 'And if I went ahead and spent a hair over $250,000, I would end up 'with a "140 MIPS, 336 Mbytes, 15.8 Gbytes" system. but *not* in one easy to use package. our users just connect to a single place i just manage a single place (sort of). my current plan is to put more cpus in the solbourne and turn the 490 into a batch machine assisted by an i860 attached processor. [neil talks about being more aggressive and getting things like aviions (or ibm powerstations i assume) and having vast quantities of mips which is interesting but might not fit the original criteria found on the keywords line above.] mips which i still feel are less usefull spread out across a bunch of small desktop machines. 'I assume that someone else is providing ethernet based terminal 'servers. yep. 'The other side of this equation is that someone else spent a 'fair amount for your serial ports, and that with 65 people banging 'away through the ethernet, you waste a lot of ethernet bandwidth and 'host CPU time with the TCP overhead. nope, not even before we switched to lat. 'Whereas if you split the system 'into smaller chunks with a bunch of serial ports directly in each of 'the Sparcstations you can optimally place the serial lines, you will 'save a noticable amount of $'s, and the hosts will be saving a 'noticable amount of CPU time. Not to mention that the users won't 'have to log in through an ethernet server and THEN rlogin into one of 'the hosts. well the serial lines have to get to the users somehow. we have a three piece campus with ~20,000 students. a lot of our access is serial via a rather expensive data switch, terminal servers are a *major* win. '>we don't like putting all our eggs in one basket. ' 'But you put it in 2 baskets, where I am proposing putting it in 6 'baskets, which is even more reliable. i was refering to vendor stability, not hardware stability. 'I'm not saying your choice is 'bad . . ., but I am saying that if you look at things in 'a different light, you perhaps could have spent about half of what you 'spent. Don't let a Sparcstation with too little memory . . . sour you on the 'possibilities . . . 'And don't get the idea that I am pushing Sparcstations, they just 'happen to be a very cost effective solution right now, plus I know 'that there will be very cheap double digit MIPS versions out from 'Taiwan soon. well this is entirely too long but . . . we gave serious thought to a variety of costs, to redundancy, reliability maintenance, ease of use and efficiency of implementation. we decided that unix timesharing on our campus was best done on larger, more centralized resources. we think multi-cpu boxes give us the best of both worlds. we have a simpler system to administer, we can incrementally increase our computer power and *all* of our users have access to the large memory, fast disks and rich software environment without having to figure out which machine to connect to. (well not entirely but almost). the biggest plus for a fully symmetric os is it gives you all these cpus and then arranges that they all get utilized without user intervention. i have evidence that the kind of computing we do here can get done with the equipment we've chosen. i don't have any evidence that if we'd spent our money (and i think it's fair to say that we support the users i was talking about on just the solbourne since i didn't include peak usage for the 490) on ~$120k worth of workstations that fact would still hold true. particularly given the fact that the solbourne appears able to support about 128 users. there's much to this but it either belongs alt.religion.computers or in our justification for purchase although it did crop up recently in comp.arch.
nash@ucselx.sdsu.edu (Ron Nash) (03/07/90)
> I am throwing this to the collective net wisdom (:-). I am asking for > suggestions and any experiences good or bad to guide us in our search. > The main uses of the computer will be teaching Unix and programming, > and research. We have Macsyma and Lisp users and neural-network > research that can be quite a drain. > > We are looking for a Unix engine that will support at least 100 > concurrent users. It would be nice if it could be expanded to support > more users if needed. The Unix should be BSD or be BSD compatible. A > major concern is compatability of BSD software. > Here is the summary of responses. Convex 1, Dec 1, Encore 1, Mips 1, Sequent's Symmetry 7, Pyramid 2, and two people are still looking. I would appreciate any additional comments. Thank all of you that responded! It has been very helpfull. -- Ron Nash San Diego State University Internet: nash@ucselx.sdsu.edu UUCP: ucsd!sdsu!ucselx!nash ############################################################################### We've been shopping recently for machines to replace our now-outdated VAX 8650's. This is what we've been looking at: Solbourne 20+ mips/processor, 4 max. Fast IPI disks on the way soon (can't give you figures, sorry). Just announced 40 mips/processor later this year. Sun 20+ mips - IPI disks now. MIPS 55 mips for their new ECL machines. Pyramid Current crop of machines based on a dual sysV/BSD universe and on the proprietary Pyramid processor with 9+ mips/processor, 14 max. Design is good, but on the way out. Other machines based on MIPS chips with 20+ mips/processor available in 4th quater. You might also want to look at HP and Silicon Graphics. Good luck. --- Name: Christopher J. Calabrese Brain loaned to: AT&T Bell Laboratories, Murray Hill, NJ att!ulysses!cjc cjc@ulysses.att.com Obligatory Quote: ``Anyone who would tell you that would also try and sell you the Brooklyn Bridge.'' ############################################################################### We are looking for the same kind of thing. At the moment, we are considering: Multiflow Trace 7/300 Encore Multimax (model 520, I think) DEC 5400 (2 of them, I think) Solbourne ??? MIPS RC6280 Sun 4/490 SGI Iris 4D (4 cpus) IBM RS/6000 530 The list is ordered from most to least BSD-like. We could live with anything above on the list above the SGI; it and IBM are too much SysV. I would be interested in any other information you can gather. We have to decide what to buy within three weeks or so. We have some benchmarks on some of these boxes. I will mail them to you if you like. -Guy Middleton, University of Waterloo gamiddleton@watmath.waterloo.edu (+1 519 885 1211 x3472) gamiddleton@watmath.uwaterloo.ca ############################################################################### Tough question, since most of the really big iron has a sysV fixation. The apps you mentioned are not floating point intensive, more on the symbolic side. I think you need something like a 4 cpu Solbourne or a ... uhm .... who is it that makes the vliw box? darn... Anyway, something like that with an FDDI link to your disk servers, say 3 or four sun4s. Put this monster in the center of your network as a cycle server and scatter smaller boxes -- sparcstation1s or something -- around for interactive junk. put the users on X terminals with their default sessions on the satellite cpus, have the cpu-intensive programs automagically run on the cycle server. -- Steve Nuchia South Coast Computing Services (713) 964-2462 "You have no scars on your face, and you cannot handle pressure." - Billy Joel ############################################################################### We have also just started a search for a similar box, so I would be very much interested in seeing a summary posted. We are currently looking at Convex (the "air-cooled Cray"). They have expandable models and are claiming POSIX compatiblity, rather than BSD or SYSV. Their compilers supposedly can vectorize and parallelize to take advantage of multiple processors. In the next week or so we will give them a number of programs/data-sets which they will run on different models with and without the vector/parallel options. Other features are high disk transfer rates and I/O processors (which they claim allows support of large numbers of interactive users). As may be obvious, the above info is from Convex's sales/marketing folks, so I am definitely interested in less biased opinions on the capabilities and reliability of this box and anything similar. -- Joe Hamlin <hamlin@blackbird.afit.af.mil> ############################################################################### Looking for a bix UNIX box ? Look at the DEC 6000 series machines. Wonderfull boxes, love them..can't say enough good about them, and I don't work for DEC. (or have much to do with them them either, just use the equipment...) Peter Theune. Johns Hopkins University, the Applied Physics Laboratory ############################################################################### You should probably speak to Encore for a configuration of that size. Boston University ... department of Computer Science has an installation fairly similar to what you describe. Encore builds MIMD parallel processors, which means they are sort of like Vaxes or other minis except that can have 2-20 CPUs (current architecture, 40 on their next system.) The CPU boards each have 2 CPUs. The company was founded by, among others, Gordon Bell who led design for, well, most of the systems DEC has produced in the last 30 years (PDP-10, PDP-11, Vax.) Rumor had it that the Encore was his design for the next generation of Vax but DEC wouldn't go for it, so he pulled Encore together. But that's just rumor. Their current system is based on the NS32532 which is about an 8 MIPS CPU, so the MIPS range is 16-160. To be honest, it's difficult to put 20 CPUs into their box although 16 should be fairly easy (the problem is just that memory goes in the same slots and there's some natural growth of memory for time-sharing systems as you add CPUs, not much of a limitation, 16*8 = 128MIPS.) Note that all MIPS ratings of course are aggregate. The system can be bought with either Mach, BSD or SYSV. I'd strongly recommend either Mach or BSD. Their Mach is being adopted for OSF's OSF/1 operating system release, so I would recommend looking very closely at Mach. They plan to move to the Motorola 88K, probably Mach will be the first release (ask them.) If I can be of any further assistance don't hesitate to get in touch. -Barry Shein Software Tool & Die | {xylogics,uunet}!world!bzs | bzs@world.std.com Purveyors to the Trade | Voice: 617-739-0202 | Login: 617-739-WRLD ############################################################################### Although neither is really available yet, I recommend the MIPS R6000-based box running 4.4BSD. Chris ############################################################################### Make sure you look at Sequent's Symmetry line. We have Dynix 3, a BSD 4.[23] port of UNIX, and Dynix/PTX, a SysV.3 version. It's all based on 80386 (ask them about 80486 if you care, though :->) processers, and the OS runs symmetric multiprocessing. The nice part about it as compared to discrete workstations of some sort is that on an idle system a single user gets the benefit of a whole lot more horsepower than any single workstation user is going to get. I can get our sales office to give you a call if you'd like. So much for my biased point of view. I hope I've at least interested you enough to take a look! Regards, Andy Valencia Dynix 3 kernel group ...!uunet!sequent!vandys ############################################################################### Take a look into getting a Sequent Symetry (one of our machines is an S27, purely development). It runs DYNIX which is sortof 4.2 level and allows you to use System V2.0 utilities if you choose. You can choose to be in either the bsd "universe" or att "universe." I have compiled a couple of large sources for bsd machines and had no difficulties, whatsoever. It is _supposed_ to be fully bsd'ish. As to power, you can put in up to twenty processor boards, each of which has two parallel 80386's. The text segment for the kernel is shared, they each have their own data segment and use semaphores and whatnot for synchronization. They are becomming quite popular with travel agencies (big, big, not boiler room, mind you). They run on a multibus, but you can use SCSI peripherials (standard if you buy it from UNISYS, I don''t know about from Sequent). It is a very fast and does not take up alot of space. The base unit is about 3.5'x3.5'x3.5'. The expansion cabinet is supposed to be a little bigger. The nifty thing is that you can let your students do driver work for a single processor environment, and then turn around and have them make the same driver work under parralel processing (they will love you for this :-) ). cbp ############################################################################### I suggest a Sequent symetry system like this one... System Configuration: type no slic flags revision MEM/1w 0 2 00000000 00.03.02 size=40.0Mb base=0x00000000 ileave-lo MEM/1w 1 20 00000000 00.03.02 size=40.0Mb base=0x00000000 ileave-hi MBAD 0 24 00000000 00.05.01 f/w version=6 MBAD 1 26 00000000 00.05.01 f/w version=6 SCED 0 28 00000000 02.12.00 ver=42 host=71 enet=080047000fa8 local ZDC 0 34 00000000 00.02.00 f/w version=14 ZDC 1 38 00000000 00.02.01 f/w version=16 ZDC 2 40 00000000 00.02.00 f/w version=14 CADM 0 0 00000000 01.03.00 sysid 0x000231 front panel type 0 PROC/386w 00000000 00.06.01 16MHz 2*32K FPA: no. 0(slic 4), 1(5), 2(8) PROC/386w 00000000 00.06.01 16MHz 2*32K FPA: no. 3(slic 9), 4(12), 5(13) PROC/386w 00000000 00.06.01 16MHz 2*32K FPA: no. 6(slic 16), 7(17), 8(18) PROC/386w 00000000 00.06.01 16MHz 2*32K FPA: no. 9(slic 19) [ Thats, 10 CPUS, 10 floating point units, 80Mb memory, 3 disk controllers ] We can put ~320 users on it all doing things.... kayessbee -- Lightning strikes! Maybe once, maybe twice. Oh, and it lights up the night! Kevin Braunsdorf, ksb@cc.purdue.edu, pur-ee!ksb, purdue!ksb ############################################################################### Look into Sequent. My company looked at them, but didn't have large enough processing needs to justify one. They are parallel Unix (Dynix) boxes capable of supporting 8 to 1000 users on from 2 to 30 80386 tightly coupled processors. What a Machine! -- Dewey Paciaffi eddjp@althea.UUCP ############################################################################### I'm not so sure a big central machine is the way to go for LISP and MacSyma users but; This machine (a Sequent) keeps about 100 rn users happy (I have yet to see it slow down). Not cutting edge and not straight BSD but it does seem to take a lot of I/O processing well. In an environment such as you describe I'd guess a cluster approach (where several identical processors share a common set of drives and each new process (or at least; each new user) gets thrown to whatever processor is least loaded at the time would be a reasonable approach. -- -dave fetrow- fetrow@bones.biostat.washington.edu dfetrow@uwalocke (bitnet) {uunet}!uw-beaver!uw-entropy!fetrow "CP/M: Remember when fast, small, useful and clean were good?" ############################################################################### Burlington Coat Factory has bought 4 Sequent Symmetry machines, and we're VERY happy with them. (We started out with 1, and liked it so much that we bought 3 more). Sequent's boxes are (coarse-grained) parallel processors. You can plug in more processor boards as system load increases. Our biggest machine has twenty '386 processors, and supports 200 users without even slowing down. (Right now we have 110 users coming in over an X.25 switch, and 30 more through telnet, and the machine is only using 20% of its processors). Parallism comes "for free" when a process forks. If you pipe the output of one program into the input of another, both programs will execute *simultaneously* on different processors. The operating system is dual-universe (like Pyramid) -- so you get your choice of BSD or SysV. The underlying kernel is BSD. We've never had any problems porting programs; they usually compile without modification. (The kernel is derived from BSD 4.2, not 4.3, if that's an issue). Call me if you have any questions you'd like to ask. -- "Live justly, love gently, walk humbly." Andy Behrens andyb@coat.com or: andyb%coat.com@dartmouth.edu or: {uunet,rutgers}!dartvax!coat.com!andyb Burlington Coat, HCI 61 Box 1B, Lebanon, N.H. 03766 (603) 448-5000 ############################################################################### You might want to check into Sequents. We have a Sequent Symmetry with over 200 logins, maybe 100 or so active at one time, plus at least one large database crunching away constantly, and the highest I've seen the load get is 2, when someone in QA fired up about 15 compiles in parallel, at the same time that someone had started up a second database. Happy computer hunting! --woodstock -- "What I like is when you're looking and thinking and looking and thinking...and suddenly you wake up." - Hobbes nhess@oracle.com or ...!uunet!oracle!nhess or (415) 598-3046 ############################################################################### Consider a system from Pyramid Technologies. We have an early 90x and it is a rock: with a single CPU configuration and about 30 users average we have had essentially no unscheduled downtime for almost three years. Their new MIServer systems can be configured with from 1 to 12 12 VUPS CPUs so the system has room to grow. The I/O controllers have a lot of smarts so they offload quite a lot of processing from the kernel. They have a "dual universe" system which is really BSD but provides a very handy and complete SysV environment (we use BSD almost exclusively except to compile SysV sources, which we then install under the BSD universe). UCSD has a bunch of them but the biggest local installation of the latest Pyramid models is a PacBell. Jerry -- / Gerald Hall, UNIX SysAdmin, (619) 587-3065 / Calma - A Division of Prime Computer Inc. / 9805 Scranton Road, San Diego, CA 92121 ############################################################################### I would suggest you take a look at machine by Pyramid Technology out of Mountain View, California, as well as those made by Sequent Systems out of Beaverton, Oregon. There are probably a few other manufacturers of such high-power, large capacity multi-user Unix systems. Maybe even the top end NCR Tower machines, and the similar top end machines from Unisys could handle that large a load, although both are System V oriented. Pyramid and Sequent both offer BSD compatible systems, I believe. I know Pyramid does. -- Chris Johnson DOMAIN: chris@c2s.mn.org Com Squared Systems, Inc. ATT: +1 612 452 9522 Mendota Heights, MN USA FAX: +1 612 452 3607 -- Ron Nash San Diego State University Internet: nash@ucselx.sdsu.edu UUCP: ucsd!sdsu!ucselx!nash