A New Distributed Approach for Building Balanced Ring for Fault Tolerance in Mesh Architecture

摘要

Fault ring (f-ring) is a popular model for the fault tolerance in grid based architectures such as 2D-mesh and torus. In the work proposed by Huxi Gu et al. [1] introduced the concept of a balanced ring(b-ring) to reduce the traffic load on the fault ring to achieve fault tolerance in a mesh architecture. However, central to their work is the formation of the balanced ring that surrounds a fault ring. In this paper, we propose a new distributed approach for the formation of the balanced ring. Our approach is based on eight-neighborhood property that requires only the local information of the faulty nodes in contrast to the global knowledge as needed by the algorithm in [1].