Software Bugs Poised to Bite as Phone System Expands
What could be more annoying than telephone conversations that keep disconnecting in mid-sentence? Or fax transmissions that end up a garbled porridge of alphanumeric gibberish? Or cellular phones that mysteriously ring and ring--with nobody at the other end?
These glitches and gremlins could become regular nuisances on the nation’s telecommunications networks, and there may be nothing we can do about them. Yes, America’s telephone systems are by far the most versatile and reliable in the world, but they are rapidly evolving into forms that make bugs inevitable. The real question is whether this network evolution will hatch little bugs that merely annoy or big bugs that can wipe out phone service for millions of people at a time.
The bugs will be a natural byproduct of new technology. In the old days, telephone switches were simply racks of electromagnetic relays routing calls from point A to point B. Nothing fancy, nothing flexible.
These days, the network and the switches that run it are digital: They rely heavily on software. These central switches, routing millions of local and long-distance calls every day, are stuffed with millions of lines of computer code. When the phone company wants to make changes in a switch, it simply reprograms it. The telephone network resembles nothing so much as a Gargantuan network of computers that happens to carry voice and data.
The beauty of these digital switch computers is that they are infinitely more versatile and flexible than analog systems. You can program them to do what you want. Think of a clever way to use telephones, and the switch can accommodate you. That’s why Pacific Bell and other phone companies can offer such nifty services as call waiting, call forwarding, call tracing, three-way calls and the bevy of 800 and 900 numbers. The catch is that, as any computernik will tell you, big computer systems have bugs. The more lines of code, the more bugs there are that can bite you.
“We have to be very concerned about bugs,” says Bruce D. Gesner, Pacific Bell’s executive director for technology introduction and support. “We go through a very, very rigorous testing of systems.” Indeed, the switches are designed to have a failure rate of less than three minutes a year. For the most part, the phone companies have done a superb job of maintaining reliability as they’ve switched over from analog to digital systems. But software is an animal that’s more hostile in the real world than in simulated tests.
“When you face a mix of traffic, you really don’t know how that will impact the switch,” Gesner acknowledges. “When you’re talking about even a single system, it’s difficult. But when we’re talking about systems of systems, then the risks are greater. All of these stored programs are interacting with each other and that makes it hellishly difficult.”
One call forward too many may cause the switch’s software to hiccup a few disconnections. A corporation piping an unanticipated mix of data and voice down the network could cause a switch to freeze up. Mind you, these are hypothetical scenarios, but software is a slavish follower of Murphy’s Law. What can go wrong, will . . . eventually.
“As the software becomes more complex, that doesn’t necessarily mean that the network has to become less reliable,” insists Ken Phillips, chairman of the Committee of Corporate Telecommunications Users, whose members include Citicorp and Hertz. “But when something does goes wrong, it will become more difficult to fix it because the code will be enormously complex.”
Exacerbating the problem is that large corporations invariably have their own complex networks attached to the public telecommunications networks. If their internal networks crash, Phillips notes, they can bring down the local switch with them. That hasn’t happened yet, but this is a contingency that local phone companies are now forced to plan for.
Past Experience
AT&T; Bell Laboratories, which defines the state of the art in digital switch software, asserts that reliability will improve. “The Malthusian story didn’t happen to us,” says Robert Carlson, executive director, digital switching systems division. “The performance over the past 20 years of the telephone network has actually improved.”
Which is absolutely true--but possibly irrelevant. When it comes to new software, the past isn’t prologue. “At a certain level of software complexity, systems can become unpredictable,” notes Bernardo Huberman, a senior scientist at Xerox’s Palo Alto Research Center exploring the dynamics of large distributed software systems. “There’s a qualitative change in behavior: You will no longer be able to talk specifics about messages going from point A to point B. Instead, you can only say that, on the average, a behavior is happening. It’s more a matter of probabilities than certainties.” In other words, extrapolating from past experience may prove more misleading than enlightening.
There’s a tremendous irony in this. At the very time that our telecommunications networks are championed as the superhighways of the Information Age, carrying data, voice, fax, etc., their performance seems destined to become less predictable and less reliable. We purchase all these terrific new services and functions at the cost of complexity and reliability. That doesn’t mean that our phone system won’t be the best in the world, it just means something we all know but refuse to admit: You don’t get something for nothing. Remember that the next time your fax gets garbled.
