Complex cascades and plectics

I should have known better. When my 1993 Pontiac TranSport van suffered a burned-out side marker lamp, I figured I'd replace the lamp in a few minutes. After all, I own a reasonably comprehensive toolkit and a shop manual for the van.

Instead, I spent almost two hours outdoors on a chilly winter afternoon dealing with a complex cascade of problems. For starters, the shop manual's illustrations and lamp-replacement procedure didn't match the actual hardware. To change the lamp, I had to displace the windshield washer tank and a headlight cluster and remove two hex nuts, one cleverly hidden in a fender pocket behind the van's battery and barely reachable with a long, thin wrench. Otherwise, to reach the offending nut, I would have had to remove a structural brace and then the battery.

Complex cascades run in two directions: In a forward cascade, a simple failure causes a branching series of events that spread unpredictably. For example, in August 2003, a couple of errant tree branches contacted high-voltage power lines, triggering a series of interlocking events and further failures that brought down the northeast's electrical grid (Ref. 1, 2).

In an inverse cascade, events in seemingly unrelated parts of a system unexpectedly interact and cause a failure. Everyone who uses a PC has encountered inexplicable operating-system glitches, many of which we can blame on cascade failures of untested combinations of programs, library routines, and device drivers.

Future automobiles will contain dozens of microprocessor-based accessories and controls, all interacting over data buses and capable of generating inverse-cascade failures. And can designers guarantee that a next-generation, 500-million gate-equivalent device or its test program will not include complex-cascade failures?

Untangling complexity cascades can consume a lot of test engineering time, and physicist Dr. Murray Gell-Mann has proposed a new branch of study called "Plectics" to explore simple and complex interactive and adaptive systems. If we're smart and lucky, Plectics will help us learn to predict— and test for— potential cascade failures. If not, our ever-more complex society can expect some interesting times.

Brad Thompson, Contributing Technical Editor, brad@tmworld.com