Event-Predicate Detection in the Debugging of Distributed Applications

Trends in the development of computer hardware are making the use of distributed systems increasingly attractive. The collection of event-trace data and the construction of process-time diagrams can provide a useful visualization tool. In practical situations, however, these diagrams are too large for users to nd them comprehensible. The ability to detect and locate arbitrary (complex) predicates within an event trace can help to alleviate this problem. This thesis enumerates ve classes of problems that a successful eventdetection strategy should be able to identify: phase transitions, mutualexclusion violations, subroutines, communication symmetry, and performance bottlenecks. Some previous e orts in this area o er an expressivity which is close to that required to meet these goals, but are hampered by an insu cient understanding of the partial order which underlies causality in a distributed-execution trace. This work de nes a partial-order precedence relationship for compound events, and extends two timestamping algorithms to support it. A new syntax for event-predicate de nition, which comes closer to ful lling the aforementioned framework than any of the previous e orts, is presented. Finally, a prototypical implementation, within Taylor's Partial-Order Event Tracer (POET), is described, issues encountered during its construction are discussed, and its performance is evaluated.