Increasingly, large parallel computing systems and networks are providing unique challenges to industry and academia in dependable computing, espe­ cially because of the higher failure rates intrinsic to these systems.