Thursday, December 30, 2004

Lessons from the airtravel debacle of Christmas 2004: the risks of efficiency

The Cincinnati Post - Comair computer crash

Air travel this past Christmas was awful. I was on a simple direct flight from NWA, but we experienced 1-3 hour delays in both directions, one from a flaw in luggage loading, the other a mechanical. Workers were burned out everywhere, from food services to flight workers -- even the toilets were trashed.

We were lucky. US Air was disabled by labor strife (I'd assume the flight attendants figure their jobs are toast anyway, so they might as well try to take out the airline). Comair was disabled by a software bug. The simple answers are "better software" and "some airlines need to go away".

But maybe there are deeper lessons to learn:
.... Tom Parsons, of Bestfares in Arlington, Texas, said the lesson he learned long ago was to avoid northern connections during the winter.

'I still can't believe what happened to Comair. You notice that Delta is sitting in the background, saying that's Comair, that's Comair,' Parsons said.

He said part of the problem is that airlines have stripped down so far that they were near 100 percent capacity for holiday rushes.

'The systems are geared to run 100 percent, and hopefully nothing goes wrong. This time, just too many things hit (Cincinnati/Northern Kentucky airport). Each one became part of the domino effect. We now know not to connect through Cincinnati in the winter or to fly Comair or U.S. Air,' Parsons said.
Maybe the deeper lesson is the risk of high-efficiency systems. Most highly efficient systems have little redundancy and unused capacity. If something goes wrong, they are prone to lockup and collapse. The same thing happened to our electrical grid in the northeast about a year ago.

In theory a system could be both adaptive and efficient -- able to run without much redundancy but also able to quickly configure to adopt to changing circumstances. In theory. In practice I work in developing complex software to augment clinical work. I worry about the risks of increasing efficiency by reducing redundancy.

I think again (and again) of an odd lecture I attended at the annual scientific assembly of the American Academy of Family Practice. I don't know how the speaker got on the schedule, his topic we rerisks associated with complex and stressed systems -- not a very clinical topic! He was inspired by a popular book of the time. I enjoyed the presentation by a fellow physician-crank. I think he (and the book he'd read) were right then, and they're right now. We pay a price by sacrificing redundancy and adaptability in favor of efficiency. It's a lesson Rumsfeld ought to have learned from Iraq (he's an idiot however, so he probably hasn't learned anything). It's a lesson we ought to learn from evolution, where highly adapted and specialized animals disappear when their ecosystem is disrupted. (Too bad this lesson is lost on the anti-evolutionists.) There are real benefits in the long term to adaptability, to excess capacity, to shock absorbers, to redundancies.

Unfortunately the market is a tool for solving local minima equations. It does not necessarily reward the ability to tolerate infrequent system shocks.

No comments: