| Summary: | 碩士 === 淡江大學 === 電機工程學系 === 89 === The mesh topology is one popular parallel network topology with high scalability and is easy to implement. A two dimensional mesh-connected network is especially desirable in use as it can be conveniently layout in the VLSI environment.
A reconfiguration approach is proposed in this paper to increase the fault tolerance for the mesh topology. The proposed approach, simple in design and efficient in implementation, is able to replace faulty nodes in the mesh network by involving spare nodes and links only, achieving desirable reconfigurability at reasonably low cost. A reconfiguration algorithm for more systematic node replacements is also provided.
Our proposed approaches involve adding redundancy (spare nodes and links) to the mesh network to upgrade fault tolerance. For improvement, we aim to simplify the above replacement operation by collocating regular nodes with a suitable number of spare nodes, and then locating the best site in the mesh network to place the spare nodes so that a satisfying replacement policy can be secured with the aid of a few spare links.
Also provided is a reconfiguration algorithm which can be used to systematically replace a faulty regular node by a spare node, able to achieve 100% reconfigurability when the number of faulty nodes is smaller than that of the spare nodes.
Compared with other proposed schemes the approach can tolerate more faulty nodes with the same overhead and no significant performance degradation.
|
|---|
