13iconNet Gain$
Back in 1983 a young CPA named John Madden came to work as information systems manager for Carsonville Metal Products in rural Michigan. He found just two computers there—an IBM mini and an Apple III.
Like most computers, they were loners. The IBM, used for accounting, didn’t pass its figures on to the Apple to be put in reports or maybe spreadsheets. The Apple didn’t talk back.
Today, though, Carsonville not only owns more machines; they’re more sociable. Its twenty-one Kaypros speak to each other through alocal area network. That’s gobbledygook for nearby computers swapping electronic files or programs—and sharing goodies like printers.
People, too, are communicating better. They needn’t swap floppy disks as often or fight their way through as much paperwork. Thanks to the computer network, Madden says, Carsonville may enjoy $1,000 a year more ineffectivework time from each office staffer. That’s not all. Sales normally are around $12 million a year; Carsonville competes in the dog-eat-dog areas of defense and auto-parts contracts; and the network may help win millions of dollars of new business. With facts handier, the executives can more easily submit low but realistic bids.
The secret is simple: The WEB network, which in late 1984 was listing for several hundred dollars per computer.
The WEB’s price sounds outrageous. It isn’t. Had Carsonville used Ethernet or another of the much-touted networks, it might have spent well over $1,000 to wire ineachcomputer—perhaps several thousand with all the trimmings included.
Instead, Madden went by the philosophy of this book. He shopped around for the least costly gizmos that could do the job well.
Souping up the IBM mini wasn’t the answer. It offered 64 megabytes of storage, the equivalent of 32,000 double-spacedtypewritten pages. But Madden worried if it could run programs fast enough with a number of people tugging at it simultaneously. “We had five terminals, and we dropped it to four,” he said, “because our mini slowed down horribly even with five.” Going the mini route, Madden would also have had to spend perhaps $30,000 on at least twenty more terminals without the brains of a full-fledged computer. In addition, he’d have had to buy new software and other hardware, so the final costs might have totaled up to $100,000. Besides, Madden cherished the notion of a “democratic” system. Machines at all locations should boast plenty of power; then people would feel more in control, and the whole system wouldn’t fall apart if a Big Brother computer broke down.
Madden might also have gone another route, buying amultiuser-systemmicro—one computer and a number of dumb terminals, somewhat like the mini arrangement.
That, too, however, wouldn’t have been “democratic.”
And with too many people piled on the supermicro, it might have run programs at the pace of a crippled snail.
UNIX, the software system developed by Bell Laboratories, was good for multiuser arrangements and allowed much more speed—but few computer and software companies had UNIX offerings out yet.
Madden, anyway, wanted his machines to run proven CP/M programs like dBASE II and WordStar; he felt they were powerful enough for the jobs he required; he needn’t mess with the more fashionable 16-bit, IBM-style micros. Hard disks would help, though. “I’d been a partner in a public accounting firm,” he said, “and I’d picked up a hard disk for myself, and it ran much faster than floppy disks, and I didn’t ever want to go back to floppies.”
So he set his sights on Kaypro 10s—with hard disks built in—and began mapping out his networking scheme.
With twenty-one Kaypros, each offering around 10 megabytes of useful storage, his company’s computers could stash away over 200 megabytes, or more than three times the space on the IBM mini.
“I don’t think people appreciate the power of micros in a network configuration,” he said.
But which network to buy?
“We were just looking at the ability to share and update the files on our disks from any location,” Madden said. “And we wanted the ability to share equipment like printers. We didn’t need an elaborate electronic mail or message system. We already had intercoms. And we also didn’t need the ability to transmit graphics.”
From the start Madden knew he hadn’t the slightest use for a deluxe network, such as one called Wangnet, which could transmitvoice and evenTVpictures but didn’t come in a version working with the Kaypro.
Well, how about Xerox’s Ethernet? Xerox was hoping that Ethernet would become an industry standard, and some users loved it. “Ethernet has been very good for us,” said a computer man with the Kentucky state government, a test user; he toldComputerworldof “excellent productivity gains.”
Ethernet, though, like Wangnet, didn’t run with the Kaypro at the time Madden was shopping, and it also had too much capacity for him.
What’s more, a version of it crashed during a demonstration when two people were trying at the same time to read an electronic file. Madden moved on. Reliability counted most.
Speed mattered, too, and he tested The WEB to see how fast it could move files from one Kaypro 10 to another.
“We set up machines A, B, and C,” said Madden. “We had machine A giving the commands for a file to be copied from machine B to machine C. At the same time, we had C transfer a file from A to B, and we had B transfer a file from A to C.
“People at all three machines hit the return keys at the same time”—to give them commands—“and then we watched what would happen.
“We found that the machines copied the files without errors in maybe one minute.” And the files were about 100K long, around fifty double-spaced typewritten pages.
Madden also tried another test to see how much The WEB would slow down the speed at which people could run programs. He compared:
1. How long a Kaypro took to sort dBASE II files electronically while not hooked up to the network.
2. How long it took while connected to the network.
3. How long a second Kaypro needed to sort the dBASE files in the first machine via the network.
“In each case,” said Madden, “there was an increase in time, but it was still acceptable.”
So Carsonville bought The WEB and saved a pile compared to doing it with an old or a new mini.
The Kaypro 10s cost $2,750 each, which, multiplied by 21, came to $55,000, and The WEB in a test version was $250 per machine for the software and a circuit board. That was just about everything but odds and ends such as $80 for 1,000 feet of telephone-style wire. Total expenses? Much less than $70,000. And even with The WEB’s list price of $350 per machine, Carsonville’scosts still wereat least$30,000 less than those of souping up the mini system or buying a new one.
Later, Madden discovered a glitch. The WEB let people flash short messages across each other’s screens, a feature called “flash,” and it crashed dBASE II.
He felt his people wouldn’t need that wrinkle, however, and they could avoid the problem just by disabling “flash.”
Ed Bigelow, president of Adevco, Inc., which made The WEB, said that if Carsonville had wanted “flash,” then Madden could have had the network or dBASE II modified to snuff out the glitch.[77]
Whatever the case, the little dBASE II problem showed the importance of prospective customers putting networks through their paces with various programs—ideally,beforebuying.
Here’s a summary of questions to ask in setting up a network:
Don’t be expensively trendy. You have just three other people in the office, and the paperwork isn’t piled that high? Then you might be much better off trading floppies.
But if you’re a busy law office, you might want to get a lot of standard, boilerplate paragraphs from a hard disk shared via a network or multiuser system.
Especially you might consider a network or multiuser system where more than one person is constantly dipping into the same data base. Suppose Production wants to know five times a day how many widget parts Inventory has left in fifteen categories. Then a network could help. The Sales Department, after all, may want to use the same computer to find out how many finished products are in stock.
Of course, if there are five hundred people dipping into a data base, a mini or mainframe would be the ticket.
Also, even Bigelow warns against bringing networks into companies in which people won’t be willing to keep their electronic files in order.
He recalls one office in which “people had been using electronic typewriters and they’d switched to micros recently and were careless about where they put their disks. They even left magnetized scissors and paper clips on them.
“People didn’t trust each other’s diskettes—or diskette habits. And on a network you can’t be sloppy. You could destroy everything if management hasn’t set up the system well. Even on some good networks people can wreak havoc on each other’s files by overcrowding disks with information. There can even be network saboteurs.
“You can’t network unless people act as a team and care about their colleagues’ records. If a company’s isn’t like that, it might be better off with a strong data-processing department to police everyone. Or they might just use micros not connected to each other—so that people will crash only their own disks.
“They might network only after they’ve successfully run employees through a training program to promote good work habits.”
You’re really planning your office, not just shopping for a connection between computers.
All kinds of questions pop up—for instance:
1. How extensive do you want your network’s file-sharing capabilities to be? Translation: “How important is it for people to share their paperwork electronically?”
2. Who’ll manage the network? Who’ll determine who can see what electronic files? Who’ll keep track of passwords? Who’ll make sure that a careless but talented worker can’t destroy irreplaceable information? You don’t, by the way, want your network manager to turn power hungry. “He shouldn’t be the network police,” jokes Bigelow. “He should be the network janitor—in the sense of keeping everything in place electronically. You might also think of him as a network teacher. He can help tell people the right way to do things.” Bigelow, in fact, warns against one person holding sway over the others by being the only one familiar with network procedures.
3. Do you want to assign special network-related duties to other people? One employee, for instance, may be the custodian ofan expensive laser printer and make sure that other people are using it most efficiently.
4. Who will work at whatnode? That’s jargon for a location orwork station.
5. Will some people share work stations? If so, you’d better decide which tasks in your office are going to be done in what corners of the room.
6. Which computers will store which electronic files?
7. How many printers and other gizmos will people share, and where will they go? Presumably, you don’t want a printer five hundred feet from the people feeding their files into it.
You also should decide how many expensive, letter-quality printers you need and how many cheapies like dot matrixes. Make up your mind—before your office wastes $4,000 on a second laser printer when just one might do.
The same idea might apply to high-speed modems if your network allows computers to share them.
8. What kinds of computers are you planning to hook up? The WEB as of mid-1984 was running only with Kaypros. But a version for IBMs was in the works.
One advantage of a big-name network, however, is that itmaymore easily get machines of different brands on speaking terms. That could count if you’re planning to trade in your old micros soon.
There’s a caveat. Some networks may work with many machines but may not be as powerful as those dedicated to one brand.
So Apples and IBMs might both share the same hard disk but might not be able to read each other’s files without costly add-ons.
It’ll depend largely on the type of network—a topic covered later in this chapter.
Wangnet will zip information over the wires many times faster than will The WEB; it’s a boon for those who need that capability but a frill for those who don’t.
Mind you, a network itself isn’t necessarily the main determinant of the speed with which files zip back and forth between machines.
The speed of your computers’ floppies in many cases will count more than the network’s transmission rate, and that’s true to a lesser extent of Winchester hard disks. Different networks, of course, may work faster or slower when teamed up with the same computer. Also, some networks seem fast when you don’t have many users. But then, with a large number, rival networks are faster.
How many computers do you want hooked in? And how much of a strain will they place on the network before it crashes or slows down to a bothersome extent?
If people are running data-base programs on colleagues’ files, a network might only work with a few users, but if they’re just occasionally swapping electronic files, the same system may accommodate hundreds of computers.
That’s especially true, say, if you’re a publishing house with many writers but few editors. The writers may not communicate much between themselves, so that the handful of editors are the only ones really taxing the system. And since they’re not running other people’s programs, even the editors won’t be that much of a burden.
The same principle might apply to an educational network in which one teacher is overseeing many students.
Are the wires and connectors strong enough mechanically, for instance? How much does the network tolerate mistakes—how crashproof is it? And what about the software? Suppose two people at once are calling up a directory to find out what’s on a common disk. Will the director accurately listeveryfile? Does the network perform well not only with the directory program but with other utility-style software in CP/M, MS-DOS, or whatever operating system you’re using?
The WEB is easy but not for theuntrainednovice. Using theKaypro version, you need to know at least the basic CP/M commands.
But once you do, you’re just about home free.
Take the PIP command, which, among other things, lets you transfer files from one disk to another on a machine by itself. To switch a file named TEST from drive A to drive B, you’d load a PIP program into your computers temporary memory. Then you might typeB:=A:TEST.
Via The WEB, though, how do you try to reach someone else’s drive?
Well, you could issue a similar command. Only, instead of sayingyouyouwanted to reach drive B, you might ask for drive H—which is how the other person’s drive might be electronically known around the office.
That’s with The WEB set up the simplest way. Even with The WEB in a more complex form, however, the procedure wouldn’t be much harder.
On the other hand, The WEB isn’t the system for you if you’re hoping just to press one or two buttons to send information on its way.
Ask, ask, ask the sales reps about what’s available.
Remember, some networks won’t let you share printers or modems or even send files to another computer. Also, will printing take longer to set up remotely than with all the equipment by your side? How easy is it?
Also, does the network havequeuing? If user A and user B both send out files to be printed on the same machine, will a disk hold user B’s file until A is done? If there’s queuing, have you given up something in return?
Find out, too, if the network isinterrupt driven? Let’s say someone’s getting information off one of your computer’s disks while you’re typing. Where should your computer concentrate its processing power? On the network? Or on your screen and keyboard? Better make it the screen and keyboard, since you want your machine to recordallthe keystrokes you’re putting in. And on an interrupt-driven system, this can happen. So there’s less chance of “Now is the time for all good men to ...” coming out as “Nwi thetme fr all good mn....”
Is error checking in use? Will the computers ask if the others received the signal okay? Will they start again if they didn’t?
Also, what about machine A using its software to work with the contents of machine B’s disks?
And can you use your pet word-processing software to print an electronic file on someone else’s printer without getting in the way of the program he’s running at the time?
What about electronic mail and equivalents of The WEB’s “flash” command? Can you easily open your electronic mailbox to see what messages are awaiting you? Will your computer even tell you on screen when you have a “letter”? Will it beep at you? If there’s a flash-style arrangement for short messages to appear on screen, can you turn it off? Not that E-Mail and “flash” are pure delights. You may not be at the screen to receive your electronic message; with a phone message or an intercom, on the other hand, the other person would know immediately that you weren’t. Then again, with E-Mail or a flash-type arrangement, you can help soften the effects of telephone tag.[78]
Also, how aboutfile locking, which keeps user A out of a file that user B is working on? That’s probably a “must.” You can’t have someone changing numbers at the same time you are and see the results add up wrong for both of you.
Another question arises. Do you want people from field offices to be able to dial in via modems? This dial-up experience—along with the general-network kind—might help you eventually make the transition to telecommuting. Obviously, however, modems may mean security problems.
Local area networks, of course, just like multiuser systems, have security risks even without modems. You’ll perhaps want to set up the network with passwords anduser-privilege levelsso that only you can get into every nook and cranny of the system—and no one can readeveryelectronic file. It’s a question of management style. In fact, in respecting people’s privacy, you might even arrange for only them to be able to read some information stored on their hard disks. At any rate, do investigate the security capability of a system very carefully before buying. Don’t let a network’s technical failings complicate office politics.
Needless to say, too, make sure that no one can find his way to the payroll data base to give himself an unofficial raise.
What Kind of Protocol Does the Network
Use?
“Protocol” is just a set of rules telling how computerlike gizmos speak to each other.
By the way, networks can share a protocol but still not be on speaking terms. The WEB uses Ethernet’s basic protocol but can’t hook up to it, since Ethernet transmits information faster across the network.
Wait. There’s one other complication. Different Ethernet-style systems—from different manufacturers—speak different dialects.
“Assume nothing,” says Bigelow on the issue of whether different networks or machines will work with each other. “Nothing’s obvious.”
Before buying, insist if possible that the sales rep set up his hardware andshowyou the “compatibility” he’s been claiming.
A WEB-type network might be a nightmare for the lazy and sloppy. It requires hooking printed circuit boards up to the right leads of chips in the Kaypro. And you must also solder the telephone-style sockets into which your computer’s network cord plugs.[79]
But a good data-processing department shouldn’t have any problems, according to The WEB’s makers.
What about small businesses without data-processing people?
They might buy both their computers and the network from a systems house if possible—a company that will do more than the average computer store in getting various machines to work well together. But you’ve already bought your computer? And a good computer store or systems house isn’t nearby? And the network maker can’t vouch for any technicians near you? Then you might avoid the networks that need soldering and other grubby work.
Even if a well-regarded store or systems house does sell you the network, see if the technicians will get it runningbeforethey set it up at your business and you officially take delivery. Better still, see if they’ll do that before you officially accept delivery.
How Much Support Will the Manufacturer
Give You?
Often, if software comes with a computer, the software house will buck you back to the computer manufacturer if you have a question or problem.
And that may or may not be true with network systems.
If the network is built into the computer, you might want to see if you can also get direct support from the company responsible for the networking system.
It works the other way, too. If you buy a network from an independent manufacturer, is the firm familiar enough with your computer to make sure the system works well with it? Know exactly who will do the servicing and consider a service contract, which usually costs 1-2 percent of the hardware costs per month.
In shopping around, you shouldn’t worry about jargon so much as you do about the performance of the network asyouperceive it.
Below, however, are some necessary terms to master. We’ll begin with three kinds oftopology—the word for the way a network is laid out:
A cable hooks up a bus network’s computers in parallel in the manner of lights on some Christmas trees.
A bus network may offer advantages in office layouts. You don’t have to clutter things up with a whole series of wires running back to fancy equipment in a central location. You just lay one main wire with sockets that the individual micros plug into.
Other equipment needsmaybe simple. The WEB, a bus network, requires printed circuit boards for all computers, installed in the machines themselves. That’s about it other than the software and the wire.
Fancier systems using a bus—like most versions of Ethernet and Corvus Systems’ popular micro network called Omninet—need afile server.
A file server can be a computer minus the keyboard and screen but with extra communications ports to help signals get in and out in a hurry. It’s connected to the hard disk, which stores and relays the electronic files that people send it. In some networks a file server can also be a regular computer simply assigned to the job.
With a server arrangement you’re always sharing files with the server rather than directly with other members of the network. The hard disk is between you and the machine you’re trying to reach.
By the way,ifeveryone must use software from the server, that means a very, very busy hard disk—and potentially slower running programs.
In star topology, the individual computers are at the points of a starlike layout of cables radiating from a file server.
If the server conks out, everyone in the network is out of luck. But then that could also happen in rare cases if a printed circuit board in a bus network became mischievous in the worst way.
Normally, computers on a star network can’t be more than perhaps 200 feet apart, and perhaps much less.
Corvus’s Constellation network uses the star.
Most multiuser systems—which I won’t call true networks—use the star arrangement to hook up the dumb terminals to their central brain.
Messages zip along a circle in one direction until they reach the right computer. If the ring’s broken, the network crashes.
Also, distance may be limited to less than a few hundred feet without costly repeaters. With the right equipment and enough money, however, you can go much farther.
One beauty of a ring arrangement is that in small networks the speeds can be very fast. Larger networks slow it down.
A ring network often usestoken passing. Think of the children’s game where, when you’re caught with a ball or other object, you’re “it.” The kids try not to be.[80]
In token passing, though, the computersdon’tdon’tmind being “it” at all. Getting the “ball” means they have the okay to send to another machine through the ring.
If you as a computer get the ball, you get the privilege of replacing it with the message you want to send. You “hold” the ball until the message you’ve sent comes back to you through thering. Then you send the ball on its way to the next computer. The whole process, of course, is almost instantaneous.
Some people say it may be easier to design software for a ring, since the token passing means there’s no chance of signals colliding.
Radio Shack’s Arcnet uses token passing.
You also may end up grappling with different wiring styles.
WEB-style networks mostly use a cable somewhat like the kind between the outside of your house and your phone jack. This cable has fourconductors—individual wires within it. Commonly, two “hot” conductors carry computer signals. Between them, often, are twoground wiresattached to the computers. You need good, solid ground connections for the networks to work right.
A twisted-pair network is what it sounds like—one normally using a pair of wires that twist around each other to form a long spiral. The twisting makes the cable less sensitive to electrical interference from radio transmitters, air conditioners, or other appliances.
You could also guard the twisted wires from electrical interference by enclosing it in woven copper or metal shielding.
Coaxial cableis a common form of shielded wire. It’s costly—it may sell for more than $1 a foot. Coaxial cable could be four times or more the cost of a twisted pair. It consists of one or more thin wires buried in plastic-type insulation under the shielding.
“Coax,” as the pros say for short, is the kind that’s normally black outside and looks like a thin snake that stretches on forever. (The pronunciation is “co-ax.”)
Not that the color’s important. “We use powder blue,” says a man with one network company.
Ethernet and Wangnet both use coax; so do cable-TVinstallations. In fact, some cable companies have transmitted computersignals. ATVcable doesn’t care if you use it for carrying a gangster movie or a bank payroll.
But just because Ethernet uses coax doesn’t mean it’s normally good enough to carry mostTV-like signals. It is justbasebandin capacity. You’d normally use Ethernet simply for computerlike messages or maybe some telephone; it’s like a single, high-speed highway.
Wangnet, however, isbroadbandand can carryTV. It resembles a whole transportation network—a highway, air corridors, and a river.
You have many well-separated channels. In fact, Wangnet is piping along signals at radio andTVfrequencies. And the right gadgetry can separate them just as easily as a good television does. You can go for miles with Wangnet—much farther than with Ethernet, which may need signal boosters after several thousand feet.
People at Wang, Xerox, and the others can get truculent and maybe even paranoid about their pet networks versus their rivals’.
“There is no light at the end of the tunnel yet,” said a friend of mine who’s a systems analyst. “Everyone is looking over everyone else’s shoulder to see what they’re doing.” And it isn’t just the computer industry. The PBX[81]makers, the switchboard manufacturers, want their shares of the action, too. This works the other way, too. IBM in 1984 said it would make Rolm—a major PBX maker—part of the Big Blue empire.
Room exists, of course, for many styles of networks—even in the same companies in some cases.
Bigelow aptly likens networks to word processors. We don’t all use WordStar; why should we all be on Ethernet?
With the right hardware, in fact, a twisted pair or maybe even a bus network might merge with the baseband and broadband ones. Networked second- or even third- or fourth-hand micros might speak to mainframes. The state of West Virginia in 1984 planned to buy hundreds of IBM computers. Many reportedly would be on Omninets. As of mid-1984 an Omninet could have only sixty-three micros on it—but why not link many Omninets to big-time networks? A man at Corvus assured me that’s exactly what his company was working on for customers like West Virginia. AT&T may have had the right idea. It planned a network to work with Ethernet and Corvus, plus the RS-232-style arrangement that you already use to hook up computers with printers and modems.
That could be just the solution for some large companies thatfeel they’re just innocent civilians in the network wars.
Meanwhile, if you’re with a small company like Carsonville Metal Products, home in on your immediate needs. Again, don’t worry about high-powered networks designed for the Fortune 500 crowd. Maybe all computers someday will work with Ethernet, say, and perhaps it’ll be just as cheap for you as a WEB-style net, but it isn’t now. Meanwhile, if an Ethernet-equipped firm wants to talk to your computers today, there’s already a network in place with fairly common technical standards: the telephone system.
Everything still sound scary? Well, just forget the jargon and simply pin down the sales reps to make sure that the network will do whatyouwant.
Hire a consultant if need be. And follow the normal rule of computer shopping and check with existing customers to see if they’re happy.
Wise network shopping, as indicated earlier, can pay off.
In mid-1984, Madden, taking advantage of The WEB, was putting the finishing touches on some software modifications. With them, whenever Carsonville made new sales or bought new supplies, he could instantly see the results on the companies’ general ledger; and just as important, he and his colleagues could easily keep up with the costs of their existing contracts. They could compile a historical record, too, a big help in planning new bids. Now that Carsonville’s computers were talking, the humans might be talking more—about the new business that the sociable machines could help bring their way.