Router-shared-pair mesh: a reconfigurable fault-tolerant network-on-chip architecture

Abstract

This paper proposes a fault tolerant scheme on Network-on-Chip-based System-on- Chip (NoC-based SoC), for problems of Isolated Processing Element (PE) and Parted Regions caused by permanent faults. The scheme is referred to as Router-Shared-Pair mesh (RSPmesh). The topology architecture of the RSPmesh uses the design that a pair of neighbouring PEs share a pair of routers, and uses MUXs to provide diversity for link-connections between routers. Topology reconfiguration algorithm and routing algorithm corresponding to the RSPmesh are also proposed. Thus, when there are faulty routers or links, RSPmesh-based NoC can be reconfigured to a new 2D mesh NoC with maybe smaller size, but regular and with no faults, and it is able to serve all healthy PEs. The RSPmesh uses no spare routers, and only makes several routers disable according to actual needs in topology reconfiguration. Evaluation and experimental results show that the proposed scheme achieves significant improvements on reliability.