The software problem you're referring to was the infamous crash of the
> AT&T long distance network in 1990 when a software upgrade was applied
> to the over 100 AT&T/Western Electric #4ESS long-haul tandems in the
> US. It brought down most if not all the #4ESS switches to a screeching
> halt for the better part of a day.
No, that's not the one I'm thinking of.
I'm very sure it was a non Western Electric switch. It was made in Plano Tx (forgot the maker's name) and it was used for local calls.
A odd sequence of errors would create a condition that was not checked and the switch would go into a loop and freeze up. This happened at the same time in a number of cities -- apparently the cause circumstances were common at a certain time of day.
I found it in the archives. Here it is:
Newsgroups: comp.dcom.telecom Date: 3 Jul 91 13:37:22 GMT Local: Wed,Jul 3 1991 9:37 am Subject: Service Outages Across the Nation
The (Newark, NJ) {Star-Ledger}, Wednesday, 3 Jul 91, p. 59
"Telephone sleuths are on the trail of mysterious service interruptions"
Washington Post Wire Service
WASHlNGTON - East Coast and West Coast, the pattern has been the same: At about 11 a.m., an entire region's telephone system collapses.
For the past six days, solving the mystery of the failing phones has become an obsession for the nation's service-conscious telephone companies. Yet despite recurring similarities and clues in the half-dozen failures to date, which have struck Washington, Los Angeles, Pittsburgh and San Francisco. the detective work remains mired in unanswered questions.
Yesterday, telephones in Pittsburgh were disabled for about two hours for the second day running, underlining the phone systems' vulnerability. The basic pattern was the same-an unexplained deluge of electronic messages shutting down a computer built by DSC Communications Corp. of Plano, Texas.
The telephone companies know that the failure is in complex electronic systems that route calls. But they cannot say why the systems are failing, why the failures are occurring within days of each other and why they all begin at the same time of day. They cannot explain why the failures occur in computers that are not linked electronically and use different versions of software, the coded instructions that tell computers how to operate.
...
Each of the afflicted machines has for some reason generated millions of maintenance messages, which normally help a computer keep track of its internal operations and communicate with others in the network. These messages generally have priority over messages that are routing calls. Too many maintenance messages meant there was no room for routing calls, and the DSC machines ceased to function. The key question, said John W. Seazholtz, Bell Atlantic vice president for technology and information services, is "why is their (DSC's) system going into overload every time we get a little rinky-dink issue that should have been automatically dealt with? The software obviously has a major problem."