mjl@ritcv.UUCP (Mike Lutz) (11/30/85)
In article <9073@ritcv.UUCP> spw2562@ritcv.UUCP (S. Wall) writes: [a followup to 275@tut.UUCP] > >We have 5 masscomps here at RIT, 3 are faculty only, and the students are >very happy about that. They'd be even happier if all 5 were faculty only, >but that would upset the faculty a great deal. etc., etc., etc. -- As Casey Stengel said: "You could look it up." Far be it from me to contradict the authoritative postulations of an undergraduate ;-), but I cannot let this pass without comment. I was in on the selection of the Masscomps, and I still think they were the right choice given the constraints we worked under (I'd have gladly taken a Cray-2, but we were a few million dollars short). While we had some problems initially, these were in part due to our own over-optimism and our unconventional use of the systems (see below). Suffice it to say that were we in the same position again, I'd still choose Masscomp. Of course, having recommended Masscomp in the first place, I have a vested interest in seeing the systems work out. You can read what follows with as many grains of salt as you like; I just ask that you apply the same saline solution to Mr. Wall's fulminations. Two years ago this month, RIT dropped a bundle of money in our laps to improve the computing environment in the School of Computer Science and Technology. The kicker was that the funds had to be spent (and the equipment had to be on site) by the end of the fiscal year (July 1). Given this, we started looking at various vendors, using the following guidelines: 1. The system had to run UNIX(tm). Essential, as we think it's the best general purpose development system around, and we're educating students who at least begin their careers as programmers. 2. The vendor had to have a product available for testing, and had to commit to delivery by July 1. 3. We wanted as much power as we could get for our money, and were willing to gamble a bit with new companies. Our 8+ years of UNIX experience seemed sufficient to get us over any rough spots. 4. The systems had to support timesharing. Much as we like workstations, the money on hand was insufficient to purchase enough single user boxes to significantly reduce the load on our main systems (at the time, 2 11/780s). We're talking about a program with ~950 undergraduates, with all but the most advanced courses filling to capacity each term. 5. The systems had to support TCP/IP over Ethernet. In retrospect, this was a gamble for us (as we hadn't installed 4.2 yet, and UUCP was our only network). We narrowed the field to 3 vendors, two supplying 11/780 class systems (or a bit more), and Masscomp with 5 workstations. We chose Masscomp, in part because they used industry-standard peripheral busses and a widely-available microprocessor. We also liked the idea of smaller systems dedicated to specific purposes (R&D, graphics, systems software courses, etc.). Well, come last fall (9/84), we ran into a lot of problems. Some were attributable to the newness of Masscomp systems in general, some were due to the fact that we were essentially the only customers using Masscomps for time sharing, and some were due to our own over optimism. Examples: NEWNESS The original Masscomp serial line multiplexor would rapidly generate spurious interrupts when connected to a long, unterminated RS232 line (antenna effects), crippling or crashing the system. We fixed this locally (by telling people "DON'T DO THAT!"). The current multiplexor handles such problems much more gracefully. TIMESHARING The initial Ethernet hardware & software would go autistic, and leave the attached system isolated until it was rebooted. No other customers had complained about this because most of the Masscomps were single user machines in a lab; if the net went down, you leaned over and hit RESET. Our machines are located far from the terminals; what is more, randomly rebooting a timesharing system can be be dangerous to your health. The current network hardware and software are quite reliable. OPTIMISM We were real enthusiastic about "servers", and put all our printers (eggs) on one node (basket). Result:, if a system's Ethernet interface went catatonic, that system's users were without a printer. We mitigated this by putting at a cheap dot-matrix printer on every system (not just the Masscomps). All in all, things were running pretty smoothly by Christmas last year. The only lingering problem is incompatibility between Masscomp's RTU (SysIII + 4.1c + a pinch of paprika) and our 4.2 Vaxen. This is more annoying than crippling, as it means we have to do some work to get 4.2 programs running. The newest RTU release purportedly addresses these problems, using dual universes (much like Pyramid). I won't know for certain a couple of weeks: contrary to Mr. Wall's assertions, V3.0 for our systems has not been shipped yet, and we do not have it in any form. In summary, Masscomp produced (and still produces) systems targeted mainly towards real time data acquisition, process control, and graphics. The characteristics of such systems also make them attractive for software development (which is why we bought them), but until now such use has been tangential to Masscomp's marketing strategy. Nonetheless, we've received prompt service and solid support when problems arose (not to say that the solutions were easy or obvious, but at least we knew the issues were being addressed). Given that a Masscomp MC500 is 50-60% of an 11/780 in our environment, and that we ended up with 5 systems for much less than the cost of an 11/780, I'd say we made the right choice. I'd have no hesitation to go with Masscomps again if they met our needs. Mike Lutz P.S. On the question of Emacs: I've only glanced at GNU Emacs, but it seems to be chock-full of 4.2isms and Vaxisms, so my bet is it would be a bear to port to almost any other hardware/software combination. -- Mike Lutz Rochester Institute of Technology, Rochester NY UUCP: {allegra,seismo}!rochester!ritcv!mjl CSNET: mjl%rit@csnet-relay.ARPA
spw2562@ritcv.UUCP (Fishhook) (12/03/85)
In article <9087@ritcv.UUCP> mjl@ritcv.UUCP (Michael Lutz) writes: >Far be it from me to contradict the authoritative postulations of an >undergraduate ;-) Haven't you heard? Undergrads know EVERYTHING!! 8-) > I was in >on the selection of the Masscomps Nothing personal intended.. > (I'd have gladly >taken a Cray-2 Now you're talking! > but we were a few million dollars short) Suprising, considering what RIT charges for tuition, etc. 8-) > You can read what >follows with as many grains of salt as you like; I just ask that you >apply the same saline solution to Mr. Wall's fulminations. Make it a REAL strong concentraion on mine.. There'd've been a lot more 8-)'s, but they'd've make the article hard to read. >1. The system had to run UNIX(tm). Essential, as we think it's the best > general purpose development system around, and we're educating students > who at least begin their careers as programmers. As an undergrad, I KNOW Unix is the best around.. 8-) >Well, come last fall (9/84), we ran into a lot of problems. I'll say.. I took graphics that quarter. I guess that's why I have such a low opinion of masscomp 500's. > contrary to Mr. Wall's >assertions, V3.0 for our systems has not been shipped yet, and we do >not have it in any form. Opps.. I was told by a faculty member that it was v3.0. Don't remember who. >Mike Lutz Rochester Institute of Technology, Rochester NY >UUCP: {allegra,seismo}!rochester!ritcv!mjl >CSNET: mjl%rit@csnet-relay.ARPA It's nice to hear from someone in the know. However, I will argue some points with you. Having used 2 of the masscomps for a total of 3 quarters, I think I can safely say I've become somewhat familiar with the system. In working on projects, and just plain digging around the systen, I've found that some of the utilities (like lpr) don't seem to work right, and judging from what's been done to fix them, never will. Also, I've found that the ethernet isn't quite as reliable as you've said it to be, and this is from the past spring/summer, which is fairly recently. I'm afraid I can't give you a lot of specifics, because I forgot as much as possible as soon as possible as soon as I got out of the courses which used the systems 8-). I will grant that it is much more reliable than that first fall (ugh!), but the crashcomp/trashcomp names have already stuck... As far as 50-60% of an 11/780, it sure doesn't look that way from the user end. *aside* Who's Mr. Wall?? Could he mean me? 8-) ============================================================================== Steve Wall @ Rochester Institute of Technology USnail: 6675 Crosby Rd, Lockport, NY 14094, USA Usenet: ..!rochester!ritcv!spw2562 (Fishhook) Unix 4.2 BSD BITNET: SPW2562@RITVAXC (Snoopy) VAX/VMS 4.2 Voice: Yell "Hey Steve!" Disclaimer: What I just said may or may not have anything to do with what I was actually thinking...