ALL CIRCUITS ARE BUSY

It started innocently enough: At 2:25 p.m. on January 22, 1990, a New York City computer on the gigantic AT&T long-distance telephone network somehow came to believe it was overloaded and started to reject phone calls. When other computers on the network tried to pick up the slack for the supposedly overburdened New York City computer, they, too, exhibited the same weird symptoms. Eventually all 114 computers on the AT&T network were affected, and long-distance callers using the AT&T system all across the United States began to receive a busy signal or to hear the now famous "All circuits are busy" message. The problem continued for over nine hours, during which time AT&T lost between $60 million and $75 million in revenues. Companies who depend heavily on long-distance phone service for telemarketing were also big losers on this day. Some companies even sent their employees home because they could not make needed calls.

And what was the cause of this problem in the nation's largest long- distance network? Engineers traced the problem to a single logic error or "bug" in the software used to route calls on the AT&T network. This particular bug, like many others, arose during an effort to improve an existing software system. All of AT&T's computers use the same software, so all were affected by the same problem in logic.

While software designers are always working to ensure that software is bug-free, this is not always possible. A program like AT&T's faulty switching system can contain a million lines of code. Even when fault-tolerant computer systems attempt to reduce runaway system failure by having modules that check on each other, the entire system can go down if all modules simultaneously suffer the same malady.

Не нашли, что искали? Воспользуйтесь поиском: