Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10.2 9/5/84; site kitty.UUCP Path: utzoo!linus!philabs!prls!amdimage!amdcad!amd!vecpyr!lll-crg!seismo!rochester!rocksanne!sunybcs!kitty!larry From: larry@kitty.UUCP (Larry Lippman) Newsgroups: net.micro Subject: Re: My interesting Experiences with the 3B2 (to say the best!) Message-ID: <271@kitty.UUCP> Date: Thu, 8-Aug-85 00:29:40 EDT Article-I.D.: kitty.271 Posted: Thu Aug 8 00:29:40 1985 Date-Received: Mon, 12-Aug-85 03:59:24 EDT References: <453@brl-tgr.ARPA> Organization: Recognition Research Corp., Clarence, NY Lines: 120 This article is long, but I hope informative to anyone interested in the 3B2: > I work in a Microcomputer Lab at Rutgers University and we have been > working with the AT&T 3B2 for roughly 2 months. At this point I wish > mention my experiences and I am curious if anyone has had the same of > troubles. > > No sooner was our 3B2 plugged in and set up when it crashed. > ... > I am coming to the point of despair with this computer. I > wish I could say something good about the 3B2. Oh, it's not that bad... But, the 3B2 has been a bittersweet experience for us. We have had our 3B2 for 10 months, and are *now* quite pleased with it. We are a R&D laboratory/consulting organization and use the 3B2 both for internal purposes, and to develop software to run both on other 3B2's and on other processors running System V. Our transition to UNIX from DEC RT & RSX and Intel RMX-8x is only a year old, and after careful evaluation we chose the 3B2 as our 'larger' application machine (as opposed to 68010-based VME bus systems for 'smaller' applications). We chose the 3B2 because: (1) it appeared to be a reliable product fully supported by the resources of AT&T, and (2) it is *extremely* price competitive with other 'big name' systems. Some of our customers are extremely sensitive to processor vendor selection and *insist* that we specify 'big name' products. Obviously AT&T qualifies in this category. We were only the second 3B2 in western upstate New York (there are now several systems in the area), so local software support was almost zero. The local AT&T repair organization stocked from day one a full set of 3B2 spares and gives use excellent service response - sometimes with a few hours. HOWEVER, the fellow who does most of the 3B2 work knows absolutely nothing about UNIX and has had almost no 3B2 training - but he DOES arrive with a full set of spare parts so that *theoretically* any hardware problem can be repaired. We are on our third motherboard and third harddisk, but that is *NOT* the result of faulty hardware. Our system came as Release 1.0 (it is now 2.2) and we were experiencing 'kernal mmu alignment errors' shortly after the 3B2 was installed. Some turkey at the Hotline thought it would be a good idea to replace the motherboard, which accounted for having motherboard #2. That did not solve the problem, so they sent us a software ports fix which promtly solved the problem. When we upgraded to Release 2.2 in May, the AT&T tech came out to install new firmware eproms on the motherboard and to add the auxiliary XM connector. While doing this, he pinched a ribbon cable against some pins on the motherboard and promptly blew it up. Enter motherboard #3. At the same time, he reversed a connector on the harddisk and blew it up. Enter harddisk #2. While spending half a day trying to format the new harddisk with his 'devtools' diskette, nothing would work, so he assumed the harddisk was faulty. Enter harddisk #3, which would still not format. The problem was finally traced to an old 'devtools' issue which was trying to format the harddisk for 10 MB instead of 32 MB. He got the proper diskette the next day, and we were up and running in minutes. Reliability-wise, the system is on 24 hours a day and has never failed. The cause for our now having motherboard #3 and harddisk #3 is solely ineptness (mostly due to lack of training) on the part of AT&T. Speaking as an engineer, from what I have been able to observe, I feel the 3B2 is extremely well-built. The documentation which came with Release 1.0 was *ABOMINABLE*! Only through the kind efforts of friends at AT&T did we get enough addditional information to be productive and (virtually) self-sufficient. The documentation which came with the Release 2.2 upgrade is infinitely better! In fact, it's really pretty good and is much better than the Unisoft documentation we use. The AT&T Support Hotline is pretty good for system administration problems, but is poor for application software (try getting a C compiler question answered!). I too cursed the RJ-45 RS-232 connectors at first, but have grown to like the little suckers. If you are having interface problems, there are *3* types of RJ-45/25-pin adapters. Make certain that you are aware of what is available and that you choose the correct one. Despite all of the negative points I have raised so far, I am really VERY PLEASED WITH THE 3B2 and highly recommend it. Why? 1. It really has been reliable from a hardware standpoint. 2. It is extremely price competitive and is compact in size. The XM units now allow significant disk expansion and provide a convenient tar. 3. In all of our application programs, the machine is fast. I mean *really* fast, despite the lack of floating point hardware (although we have written code to compensate for this). I have run (and have had run for me) extensive benchmark tests comparing the 3B2 to other UNIX systems and it compares quite favorably (except for floating point - which is terrible) with other 32-bit machines. We have 30K+ lines of Fortran (much of which ported from DEC and Intel Fortran) and 10K+ lines of C running virtually bug-free. Sure, that did not happen overnight, and sure, we made a lot of mistakes and learned things the hard way, but it was a worthwhile experience. We are trying to write as much new application code as possible in C, and while there are many things yet to learn, we feel that we are in the right direction. 4. The system administration functions are rather well done using menus, which should vastly simplify things for people just learning UNIX. I personally don't use the menu SA functions much, but that is because I already learned the SA stuff and feel more comfortable manually editing files. 5. Some *terrible* bugs in Release 1.0 (lp, uucp, etc.) are now gone with the introduction of Release 2.2. Sure, we have new bugs - like with 9,600 baud direct links - but they only affect a small part of the potential 3B2 users. 6. As far as I can ascertain, AT&T is genuinely *trying* to provide better support for the product, and I believe they will eventually succeed at reaching a support level which is acceptable to most people. In my opinion, AT&T's biggest problem is that they had to find jobs for a lot of ex-telephone company people with NO computer experience: enter the reason for most of the present ineptness. At the present moment, organizations like ourselves have to grin and bear it until AT&T gets their act together. We have survived, however. There is no better way to learn something then to be forced to figure it out yourself! :-) I do not believe we are much different that others who are first on the block with a new product. +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ | Larry Lippman @ Recognition Research Corp., Clarence, New York | | UUCP {decvax,dual,rocksanne,rocksvax,watmath}!sunybcs!kitty!larry | | {rice,shell}!baylor!/ | | VOICE 716/741-9185 syr!buf!/ | | TELEX {via WUI} 69-71461 ansbak: ELGECOMCLR | | | | "Have you hugged your cat today?" | +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++