A fault-tolerant connected-cycle-based mesh architecture

摘要

Four consecutive nodes in a conventional mesh structure are first organized as a connected-cycle module. And then, multiple connected-cycle modules are combined with multiple bus sets and spare nodes to form the fault tolerant connected-cycle-based mesh architecture (FT-CCBM). With this architecture, two reconfiguration schemes, both based on the modular block, are developed. Both schemes can eliminate the spare substitution domino effect during the reconfiguration process. Simulation results show that it appears to be better to implement the FT-CCBM with 3 or 4 bus sets, and that both schemes provide for an increase in reliability over the interstitial redundancy scheme (Singh, 1988) and the multi-level fault tolerance mesh (MFTM) (Hwang, 1996), at the same redundant spare ratio. Especially, with global reconfiguration, the reliability improvement ratios per spare (RIPS) can be at least twice of that of the MFTM scheme. Furthermore, the lower port complexity in spare nodes as compared to those in both of the aforementioned schemes, and versatility in reconfiguration capability are two additional merits of the proposed architecture.