Defcon - Preventing Overload with Graceful Feature Degradation

Explore a 15-minute conference talk from OSDI '23 that introduces Defcon, an innovative system designed to enhance the reliability of large-scale, globally-distributed Internet services. Learn how Defcon employs graceful feature degradation to maintain service availability during capacity outages and unexpected resource demands. Discover the common interface Defcon provides for product developers to define feature knobs, enabling gradual disabling of less-critical features to reduce resource consumption. Understand how Defcon automatically tests these knobs to evaluate product- and infrastructure-level trade-offs. Gain insights into Meta's real-world application of Defcon to improve global product availability during worldwide demand surges and large-scale infrastructure failures.