A Taxonomy of Distributed Debuggers Based on Execution Replay.

Abstract

This paper presents a taxonomy of parallel and distributed debuggers based on execution replay. Programming of distributed and parallel systems is a complex task. Amongst the many factors contributing to this complexity, the nondeterminacy of these systems is an important one. Execution replay is a technique developed to facilitate the debugging of nondeterministic programs. Execution replay has very broad applications and not every algorithm is applicable in every situation. This taxonomy provides a precise classification of replay debuggers using nine criteria. From this classification, it is easier to determine a debugger's scope of application, outline its strengths and weaknesses and compare it with others. This taxonomy is illustrated and validated using a collection of existing replay debuggers. Keywords: debugging, nondeterminism, execution replay 1 Introduction It is well known that programming of distributed and parallel applications is a complex task. Furthermore, very few ...