Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10.2 9/18/84 exptools; site whuxlm.UUCP Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxb!mhuxn!mhuxm!mhuxj!houxm!whuxlm!sdb From: sdb@whuxlm.UUCP (Brener Stanley) Newsgroups: net.invest Subject: Re: Tandem Message-ID: <677@whuxlm.UUCP> Date: Wed, 13-Feb-85 13:30:05 EST Article-I.D.: whuxlm.677 Posted: Wed Feb 13 13:30:05 1985 Date-Received: Thu, 14-Feb-85 02:28:57 EST References: <380@decwrl.UUCP> Organization: AT&T Bell Laboratories, Whippany Lines: 59 > >Newsgroups: net.invest > >Path: decwrl!decvax!bellcore!allegra!ulysses!mhuxr!mhuxj!houxm!hou2h!geopi > >Subject: Where is TNDM going? > >Posted: Wed Jan 2 12:17:44 1985 > > > >Please, I am soliciting opinion, prognostications, clear visions, > >expert analysis, rumor, even idle talk concerning Tandem Computers' > >future in the fault-tolerant marketplace, and their capacity for > >continued growth, given their various emerging competition (Synapse, Stratus), > >and the announcement last year of the AT&T 3B20D. > > >- George P. Cotsonas > > AT&T Consumer Products Laboratories > > Holmdel, New Jersey > > ...houxm!hou2h!geopi > > > I used a Tandem NonStop system for about 1.5 years. While conceptually a > nice machine I did not really care for it. The operating system was a real > pain to use, and the software in general was extremely hoggish. In > general, the only way to solve performance problems was to buy more memeory, > CPU's, and disk. This makes a relatively inexpensive system awfully > expensive real quick. Hardware fault-tolerance seemed to work, but a few times > I saw the whole system crash. Software fault-tolerance is acheived by > checkpointing, is not user transparent, and must be debugged by crashing CPU's. > For any reasonably complex application it was practically impossible to > implement correctly. For all of the above reasons I would think longer > and hard before I would be convinced that a Tandem system is the way to > go. > > > Although I am not intimately familar with them, it is my impression (possibly > incorrect) that all of the other vendors, Stratus, Synapse, AT & T, and > Sequoia (you forgot this one), provide software fault-tolerance transparently > to the user. This fact alone makes them more desireable than Tandem. Tandem, > however, clearly has the advantage right now of market share, and that alone > may be enough to keep them on top for a very long time. > > > Ken Wilner > > decvax!decwrl!dec-rhea!dec-quill!wilner I worked on a Tandem for about two years. As Mr. Wilner states the software checkpointing can be a real pain for a complicated application. However, Tandem has an excellent interprocess communication facility. They support many communications protocols. These protocols simplify communications programming. The basic routines used are just read write, control, and setmode. In fact, Tandem machines are often referred to as communications boxes. Their file system supports record locks,keyed files, files with secondary keys, files partitioned across many disks, and simpler kinds of files. They have excellent networking among a group of Tandem machines. Using their networking package I often forgot which system I was logged onto. Even without the fault tolerance (ie mirrored disks etc.) I think that Tandems are really great machines. SD BRENER