Fault tolerant matrix operations for networks of workstations using multiple checkpointing
Youngbae Kim, James S. Plank, Jack J. Dongarra · 2002
Recently, an algorithm-based approach using diskless checkpointing has been developed to provide fault tolerance for high-performance matrix operations. With this approach, fault tolerance is incorporated into the matrix operations, making them resilient to any single process failure with low overhead. In this paper, we present a technique called multiple checkpointing that enables the matrix operations to tolerate a certain set of multiple processor failures by adding multiple checkpointing processors. Results of implementing this technique on a network of workstations show improvement in both the reliability of the computation and the performance of checkpointing.