2000
How Complex Systems Fail
Eighteen truths that explain every outage, crash, and disaster you will ever be part of.
“Complex systems run in degraded mode.”
Failure is ordinary
Cook writes for medicine and operations, but software engineers recognize the pattern. Systems drift into unsafe states while appearing fine. Redundancy adds coupling. Postmortems find "human error" because the graph is too large to hold in one head.
Degraded success
Most outages are recovered, not prevented. Success comes from people compensating for brittle automation. The system "works" until it does not.
Required reading for builders
Short enough to read in one sitting. Pairs with any paper that promises reliable AI in production. Complexity does not negotiate.
- collection
Why big systems behave in ways none of their parts intend — and how they break.
← previous · 1998
The Anatomy of a Large-Scale Hypertextual Web Search Engine
next · 2005 →
Power Laws, Pareto Distributions and Zipf's Law