When the Internet Collapsed: The 40-Year Evolution of TCP Congestion Control
In October 1986, something alarming happened on the Internet. Data throughput between Lawrence Berkeley Laboratory and UC Berkeley—sites separated by just 400 yards and two network hops—dropped from 32 Kbps to 40 bps. That is not a typo. The throughput collapsed by a factor of 1000. The Internet was experiencing its first “congestion collapse,” and nobody knew how to fix it. Van Jacobson, then at Lawrence Berkeley Laboratory, became fascinated by this catastrophic failure. His investigation led to a landmark 1988 paper titled “Congestion Avoidance and Control,” which introduced the fundamental algorithms that still govern how data flows through the Internet today. The story of TCP congestion control—from those desperate early fixes to modern algorithms like CUBIC and BBR—is really a story about how we learned to share a finite resource without a central coordinator. ...