So we hear you like papers
summary
Paper introduces unreliable failure detectors as an abstraction that decouples failure detection from implementation, defines two properties—completeness (strong vs weak) and accuracy (from perfect to eventually weak)—and proves that even detectors that make bounded mistakes can solve consensus despite FLP. The claim makes consensus practically implementable: implement a detector meeting the chosen completeness/accuracy bounds, integrate it into a consensus algorithm, and the system reaches agreement after periods of instability.
AI-written summary · about
related talks
48mFault Tolerance on the Cheap Making Systems That Probably Won't Fall Over
54mSeeing Like a Programmer - Resiliency, Limits, And Moral Hazards In SE
49mBeware of Survivorship Bias!
32mBoF: Quality and Testing Processes for Zephyr + Discussion
01h 30mContracts, Safety, and the Art of Cat Herding
34mWhy Good Software Goes Bad
49mHow our systems gain resilience, and why failures are useful
01h 05mKeynote - Without Resilience, Nothing Else Matters