Online Fingerpointing: Just-in-Time Problem Diagnosis for Distributed Systems

Distributed systems are growing both in size and complexity. In the event of a system failure, this makes it increasingly difficult for systems adminis- trators to determine which component failed. Exist- ing tools and algorithms have been designed to di- agnose problems, but they rely on offline analysis. This work explores the possibility of online failure diagnosis that operates as the distributed system un- der observation is running. A framework for online fingerpointing is presented and evaluated.