David N. Dobrin
When the idea of a microcomputer lab was first conceived, we all had a pretty clear idea of what such a lab would be, essentially, 25 PCs in a room. Originally, the PCs would have been IBMs or Apple IIs. Now they can be Macintoshes, IBM ATs, Tandys, Compaqs, or Zeniths, as well as IBMs or Apples. Even still, they are PCs. Each PC would have only one user. If you wanted to move a program or a paper from one machine to another, you would have to move a disk. And, except in the rare, wealthy university, each PC would have only floppy disk drives, so that all data manipulation was driven by floppies.
The reasons the PCs were "configured" this way (if so fancy a word can be used for so simple an idea) were threefold. First, the system was (relatively) cheap. Second, there were (essentially) no maintenance headaches. Third, the PCs were under the control of the writing lab. The English department could pay for the PCs with a relatively small capital outlay; it wouldn't be handing over juice (read connection fees) and control to the computer science department for the rest of time.
p. 8
In this article, I suggest that the economics of buying microcomputer labs have changed, and with the economics, the reasoning. The main point is that the initial capital outlay for a minicomputer lab is only slightly greater than that for a microcomputer lab. And the minicomputer gives you much, much more. The operating system and software that come with the minicomputer give you more computer power than you can get with the PC. More importantly, the minicomputer is more pedagogically effective because it allows you to network. In the bad old days, minicomputers broke down a lot and were inextricably associated with the computer science department. Now, however, they break down much less than PCs and require no great expertise to purchase and install. You just stick them in a room, like a PC. Maintenance is more of a headache, but nothing that a trained student can't do.
The advantage of networking can be put very simply: it allows you to transfer files between computers electronically. This ability may not seem like such a big advantage-after all, you can always pick up a floppy and move it to another machine-but experience shows that it is indeed an advantage.
For example, hooking the computers together allows them to share resources. Disk storage is one such resource. With 25 PCs standing alone, disk storage requires 25 small hard disks; when the PCs are connected, they can all share one hard disk. The resulting cost difference is enormous. For most schools, it is economically impractical to put hard disks on each stand-alone PC. For a networked system, however, a single large hard disk is perfectly affordable.
Once the hard disk is shared, the PCs can also share software resources. Word processing on 25 PCs requires 25 copies of a word-processing program. On a networked PC, only one copy is required; however, the cost-saving is not 25:1, because most companies have per-user licensing agreements. The savings are substantial, nevertheless.
No licensing agreement covers the most important shared resource: printers. Printers are a big problem in a microcomputer
p. 9
lab. In the ideal writing lab, each user would have a printer available; an important part of writing on a computer is the ability to produce printouts at will. Good printers, however, can cost as much as computers; usually, the best that the lab can do is to put one low-quality dot-matrix printer for every two stand-alone PCs. The output from these printers is poor; printing out is slow and ties up the machine; and when 12 students are printing at once, the noise is incredible. With a network, all the students can print their papers out on a fast, high-quality, quiet laser printer at about half the cost of 12 Epson printers. Printing out, moreover, doesn't tie up the computer because the computer puts the files in a special queue, called a "print spooler."
The other advantage of networking is less economically salient, but more important. The basic educational advantage of writing on a computer is that it makes writing and revising easier. Networking significantly enhances this advantage by making papers easier to read and edit. Networking also makes the mechanics of teaching somewhat easier and, most importantly, allows the teacher to get closer to the students during the writing process. Consider the possibilities in the following areas
As many commentators have noted, one effect of teaching writing in a computer lab is to make the teacher and the student co-conspirators rather than enemies. Teacher and student help each other make a paper; without the lab, the teacher can easily be cast in the role of destroyer, ruining what the student has created. A networked lab brings the teacher and student closer to that kind of relationship.
I hate to say this about a product I use every day, but DOS is an inferior operating system. It was designed for a single-user computer, and it does the minimum possible even for that. With a minicomputer, you don't have to use D0S. Colleagues in an educational institution can license Berkeley UNIX for very little, and Berkeley UNIX is a very good operating system. It was designed for networked computers in a college environment, and the list of things it does is genuinely remarkable.
File Sharing Berkeley UNIX allows you to read other people's files or execute other people's programs as if they were your own, as long as permission is given. UNIX also provides full mail and message facilities. You can talk (type) to another terminal. You can (with permission) look at another person's terminal session. A full computer bulletin board lets you post interesting
p. 11
things, assignments, questions, appointments, etc. UNIX even has an alarm clock.
Writing Tools An extremely sophisticated text editor (called VI), formatter (called TROFF), and even parts of the vastly overrated WRITER'S WORKBENCH come as part of the operating system.
Print Spooler The Berkeley UNIX has an excellent one.
File-Handling Tools You can search one or more files for one or more pieces of text. You can take one or more operating system operations and chain ("pipe") them or redirect the output. You can, for instance, create a file with every occurrence of the word "really," in all your students' papers, format the file, print it out, and mail it to all the students with one string of commands.
Development Tools A sophisticated "C" compiler and program development facilities come along with the operating system. The system also includes FORTRAN and PASCAL compilers.
Other Tools Berkeley UNIX is nice to work with because it lets you run as many programs as you want, all at the same time. You can, if you want, read a paper, balance your checkbook, and play games (Ssh, don't tell the students) just by switching from program to program.
Berkeley UNIX was developed by programmers for programmers, so I'm just scratching the surface of its capabilities. Most of the really fancy capabilities require a lot of study in order to use.
p. 12
What I'm talking about so far can be picked up in one or two days.
Cost comparisons are very tricky because extra functionality in some systems adds very little to the price tag. I am choosing to compare the following two systems, which have, in my view, the minimum necessary functionality.
Computers: 25 IBM PCs with two floppy disk drives 3-COM Network Server, which runs the system Memory: 640K RAM PCs; server 8 MB 70 MB hard disk in the server Operating System: DOS Accessories: Print Spooler, modem Word-Processing Program:WORD PERFECT Table 1: System I Computer: Data General Bulldog Memory: 10 MB RAM 120 MB hard disk Operating System: Berkeley UNIX 4.3 BSD Accessories: Print Spooler, modem Word-Processing Program: WORD PERFECT Table 2: System II
p. 13
From the point of view of the user, these systems are identical. Each allows the user to log on, use the word-processing program, store the text on a hard disk, print text out, and send text to other people.
The PC system is somewhat slower and can accommodate fewer users because it has less memory.
25 IBM PCs $40,500 25 3-COM Boards 15,000 1 3-COM file server 6,000 Memory 8 MB Disk Memory 70 MB DOS 1 Networking Software 4,300 1 Modem plus software 900 1 WORD PERFECT 4,000 25 Monitors and cards 13,000 25 640K cards 10,000 Cables & Installation 2,000 *Total: $95,200 (-20% disc.) 76,160 *Source = MBI Computers, Inc. 1 Bulldog $40,000 24 I/O ports plus console Disk Memory 120 MB 1 year maintenance Berkeley UNIX 4.3 BSD Print Spooler Modem 1 WORD PERFECT 4,000 24 Wyse terminals @ $500.00 12,000 Cables and installation 2,000 *Total: $58,000 Source = Data General Corporation Table 3: Cost Comparison
p. 14
In the UNIX network, WORD PERFECT is included for comparison purposes, but it is not recommended. The UNIX text editor, VI, is free. If you want a word-processing program rather than a text editor, better and cheaper programs are available for UNIX.
The boards and networking software are required in each PC before the PCs can be connected to each other. I am assuming the standard IBM volume discount on the purchase of computers.
Essentially, the cost of computers is the same in each system, but in one, networking is free, and in the other, it costs $25,000.
A Bulldog is a minicomputer made by Data General. There are many equivalents. DEC (Digital Equipment Corporation) makes the best-known minicomputers, the VAX line. Prime Computers, Hewlett Packard, and many others also make good minicomputers. I pick the Bulldog for comparison because it is slightly faster and cheaper than a VAX, but is still made by a Fortune 500 company. Data General, in my experience, is much easier to deal with than DEC, as you will read later.
A Bulldog is about the size of a small refrigerator; it fits in the corner of an ordinary classroom. No air-conditioning is required, though if the temperature got above 90 degrees, the Bulldog, like an IBM PC, would probably get cranky. The memory and disk storage are more than sufficient for 25 users. At full utilization, depending on the kinds of programs being run, each user should find the Bulldog about as fast as a PC with a hard disk. The storage memory allows each of 120 users to have 500K (enough to hold a book manuscript) of permanent memory.
To go above 25 terminals, you need to add boards called terminal multiplexers (into which you plug the terminals) and some memory, so that performance won't degrade. The upper limit is probably around 30 terminals.
The per-user cost of the Bulldog, including bells and whistles, is $2320. How much less are 25 PCs in a room? The answer depends largely on how much you would want resources shared by users of a minicomputer.
Let's try to calculate the cost of the cheapest possible system. According to PC Magazine, a PC called PC+ seems to work fine. It
p. 15
retails (2 disk drives, 640K, monitor) for $725.00. Even including the inevitable extras (delivery, DOS, cables), that's 25 PCs for $20,000, or $800.00 per user.
Now we need software. PC WRITE is $10.00 per user. But let's assume that some of your users will want a relatively sophisticated program. For comparison purposes, let's add WORD PERFECT ($239.00) and bring up the user cost to over $1000.00. Now let's add a thesaurus ($70.00), a spelling checker ($60.00), an outliner ($70.00), and some utilities and games ($35.00). The cost is now $1275.00 per user. Still, not bad. Now, let's add printers. The cost of 12 printers is $6000.00; the cost of a laser printer is around $2000.00. The cost difference is $4000.00 or $160.00 per user. With better quality or more printers, the difference is even greater. We're now at $1435.00 per user.
The difference of $800.00 is substantial, but so is what you get for the extra $800.00. Are there other expenses for the stand-alone PCs? That depends on whether you want to extend the system. If you want to teach scientific and technical communication, you need a better word-processing program. MANUSCRIPT, the current program from Lotus, costs $500.00. Suddenly the difference is minimal. If you want to change word-processing programs, add graphics capabilities, or add other programs that appeal to you, the difference goes down again. Don't forget; to keep costs down, we are using a clone. Clones are never truly compatible. Software distributors don't return your money when their software doesn't happen to work on your computer. Budget at least $300.00 for compatibility losses. Suddenly the difference is negative.
If you ever did want to network this system, your best bet would be to buy a Bulldog, and use any of the several free terminal emulation programs to make these computers into terminals. (You can upload or download with a free program called KERMIT .) In such a case, of course, the difference is substantially negative.
Acquiring a minicomputer system is a problem. Buying a minicomputer is no easy task; in part, because hardware and software questions are complex; and in part, because computer companies make it difficult.
p. 16
Consider the problems I had in writing this article. I wanted to quote Digital's price for a writing lab, because Digital's VAX computer is the industry standard.
Innocently, I called Digital and asked how much a VAX would cost. They wouldn't tell me. (They don't approve of comparative shopping.) Three people promised to send me a price catalog (very difficult to figure out, because of the complications), but none of them actually did.
I finally reached the person who is in charge of selling writing labs at Digital. I explained what I wanted. She said that she couldn't give me a price over the phone, but she would send me a fully worked-out quotation. She made the promise twice, but never delivered.
"But," she said in a grave tone, "we certainly don't recommend the system you describe. We prefer to run VMS (a Digital proprietary operating system) and Digital's own word-processing system. With it, you get a thesaurus. And if you do want UNlX, we recommend Digital's UNIX, called ULTRIX."
This is, frankly, appalling. Berkeley UNIX is essentially free to educational institutions; ULTRIX and VMS emphatically are not. ULTRIX has a reputation for being buggy and slow; the reputation seems to be changing, but I still wouldn't trust it. VMS was not designed for the college environment, offers one-tenth the software that Berkeley UNlX provides free of charge, and doesn't have anywhere near the networking capabilities that UNIX has. And Digital's word-processing program is much inferior to Bulldog's VI. It's slower, less functional, and requires a special template for the keys. I have used both word-processing programs.
I then called Data General. They gave me the price over the phone and didn't try to sell me something I didn't want. I had worked with Data General before and may be giving them too much benefit of the doubt. But they do seem to be easier to work with.
I am not, however, recommending any product or company mentioned here (except Berkeley 4.3 BSD UNIX); there are many other minicomputer companies, and they all, I'm sure, have some merit. If my experience with Digital is typical, then buying a minicomputer is like buying a very expensive used car. Where can you get help? One possibility is the computer science department at your school. Unfortunately, unless those in the computer science department are really unusual, they will have opinions. Usually those opinions won't adequately take into account your needs for a writing lab. Do not let them take over the process for you. This
p. 17
aspect may remain the one advantage of buying single PCs; the computer science department doesn't think it has propriety rights over their purchase.
One other thing, a matter of terminology: You should be buying a minicomputer, not a mainframe computer. Using the word mainframe in a purchasing conversation is like telling that used-car salesman that you don't know much about cars, but are relying on him for good advice.
I'm saving the pill until last. Maintenance of a networked system is essential. People, that is, have to be given access to the system and given storage space. Work must be backed up periodically. Occasionally, even in the best system, problems arise that must be corrected. New programs have to be installed from time to time. All these procedures can amount to as much as five hours of work per week. Trained students can perform maintenance easily enough; there are plenty of books on how to administer and maintain a UNIX system. Speaking of reliability, one last thing. Many people think that big computers are less reliable than PCs.
On a frequency basis, this is clearly not true. APC "crashes" (hangs, so that you must re-boot the system) as often as once or twice a session, far more often than a VAX. The trouble is that when VAXs do crash, they put a lot more people out of work while they are down. Fortunately, they don't crash very often, and when they do, they solve the problem themselves and come back up 95% of the time. MIT, for instance, has about 200 VAXs (or VAXen, as the hackers call them) running 24 hours a day, at the moment of this writing. None has been down in the last day; the average time since the last crash was seven days, and the crashes lasted no more than 15 minutes. This information was gained, by the way, in about 5 minutes.
This article is a pitch for minicomputers, I admit, but it is a disinterested pitch. I don't sell minicomputers. I write this article because I've used both, and I found life much better and easier with minicomputers. Good luck.
David Dobrin currently serves as a consultant with Lexicom in Cambridge, Massachusetts.