Compiler assisted fault detection for distributed-memory systems
Chaohui Gong, Rami Melhem, Rajiv Gupta · 2002
Distributed-memory systems provide the most promising performance to cost ratio for multiprocessor computers due to their scalability. However the issues of fault detection and fault tolerance are critical in such systems since the probability of having faulty components increases with the number of processors. We propose a methodology for fault detection through compiler support. More specifically, we augment the single-program multiple-data (SPMD) execution model to duplicate selected data items in such a way that during execution, whenever a value of a duplicated data is computed, the owners of the data are tested. The proposed compiler assisted fault detection technique does not require any specialized hardware and allows for a selective choice of redundancy at compile time.>