A Fault-Tolerant Adaptive and Minimal Routing Approach in n-D Meshes

In this paper a sufficient condition is given for minimal routing in n-dimensional (n-D) meshes with faulty nodes contained in a set of disjoint fault regions. It is based on an early work of the author on minimal routing in low dimension meshes (such as 2-D meshes with faulty blocks). Unlike many traditional models that assume all the nodes know global fault distribution, our approach is based on the concept of limited global fault information. First, a fault model called fault region is used in which all faulty nodes in the system are contained in a set of disjoint regions. Fault information is coded in a 2n-tuple called extended safety level associated with each node of an n-D mesh to support minimal routing. Specifically, we study the existence of minimal paths at a given source node, limited distribution of fault information, minimal routing, and deadlock-free routing. Our results show that any minimal routing that is partially adaptive can still be applied as long as the desti...

• Type: Conference
• Year: 2000
• Where: ICPP
• Authors: Jie Wu
• Keywords: Distributed And Parallel Computing, Fault Information, Fault Region, ICPP 2000, Minimal Routing