18th International Parallel and Distributed Processing Symposium (IPDPS'04) - Workshop 7
Fault Tolerance and Scalability of the Reconfigurable Mesh
Santa Fe, New Mexico
April 26-April 30
ISBN: 0-7695-2132-0
This paper considers fault-tolerance on the R-Mesh and LR-Mesh models. We propose a technique to identify a healthy sub-mesh from a faulty model using the removal fault model. Then, we use scalable algorithms to simulate the faulty model on the resulting healthy sub-mesh. We also extend this work to cover more restrictive variations of the reconfigurable mesh, specifically, the NXR-Mesh and NXLR-Mesh. The overhead for the R-Mesh and NXR-Mesh is O(log n), and we obtain a constant overhead for the LR-Mesh and NXLR-Mesh.
Citation:
Alejandro Estrella-Balderrama, Jos? Alberto Fern?ndez-Zepeda, Anu G. Bourgeois, "Fault Tolerance and Scalability of the Reconfigurable Mesh," ipdps, vol. 8, pp.172b, 18th International Parallel and Distributed Processing Symposium (IPDPS'04) - Workshop 7, 2004