[comp.dcom.telecom] How the AT&T Accident Snowballed

CAPEK%YKTVMT.BITNET@cunyvm.cuny.edu (Peter G. Capek) (01/18/91)

An article in the {New York Times}, Jan 14, 1991 gives some
information about the cutting of the fiber cable in Newark on January
4.  I'll outline the highlights:

- Cable through Newark was one of AT&T's ten busiest.

- Employee was in manhole removing an outdated (copper?) cable, but
failed to read the label on the one he cut, so didn't realize it was
the wrong one.  His supervisor also failed to notice the error.
Further, a second crew which was supposed to have marked the cable to
be cut never arrived at the site.  Had they been present, they
presumably would have averted the error, and if not, at least realized
promptly what happened.  As it happened, the cut was not located for
over an hour.  (The article implies that it was not apparent that the
cut cable was fiber rather than copper, because the cutter used tended
to crimp the end, and because the "weak" laser light which was being
transmitted was not visible to the naked eye.)

- AT&T Operations Center in Bedminster, NJ was not notified that work
was being done in the proximity of the cable, so when they discovered
the failure, had no idea where to send repair trucks.  So they had to
resort to a time-domain reflectometer ("a sonar-like device") to find
the point of the break, a procedure which took between 60 and 90
minutes.  Apparently, the procedure of notifying the Operations Center
about work in progress near its cables was routinely done, but is no
longer done, in an effort to reduce costs.

- The computers which reroute traffic in such situations had not been
"programmed" to give priority to the air traffic control (leased
circuits, I presume).  So, even though routine voice traffic was being
rerouted within fifteen minutes, New York's three airports lost
contact with their long-range radar for 102 minutes, delaying and
cancelling many flights.  But the F.A.A. had not paid for any special
rerouting service.

- New York City accounts for 1/5 of AT&T's domestic calls and 1/3 of
its international calls.

- The problem occurred despite a warning of vulnerability it received
on Nov 18, 1988, when a construction accident caused the same cable to
be cut twenty miles to the south.

- Following that, the traffic on the fiber was reduced, and
constructed alternates.  But there's still not enough alternate
capacity to assume the load in the case of an outage such as this one.

- 3/5 of all calls attempted to and from New York City during business
hours on Jan 4 were not completed.  (No statement is made about how
much greater the number of attempts was than is usual for a Friday.)
ATM lines and leased lines servicing the financial industry were also
affected.

- The cut occurred in a personhole at a point where a lot of traffic
is funneled together to cross the Hudson River into New York.
Unfortunately, the hole flooded and pumps had to be brought in.
Further, there was so little slack in the cable that a new section had
to be spliced in, thus doubling the number of connections to be made.

- The incident made at least one customer look to Sprint for service.
It isn't clear whether that unnamed "medium-sized business" will move
to Sprint entirely, or use both carriers.

- Donald E. Lively, a retired AT&T "expert on restoring service"
couldn't understand why the problem wasn't fixed within half an hour
or so.  As it was, it took over seven hours.


Peter Capek