Adaptive Fault-Tolerant Deadlock-Free Routing in Meshes and Hypercubes

We present an adaptive deadlock-free routing algorithm which decomposes a given network into two virtual interconnection networks, V IN 1 and V IN 2. V IN 1 supports deterministic deadlock-free routing, and V IN 2 supports fully-adaptive routing. Whenever a channel in V IN 1 or V IN 2 is available, it can be used to route a message. Each node is identiied to be in one of three states: safe, unsafe, and faulty. The unsafe state is used for deadlock-free routing, and an unsafe node can still send and receive messages. When nodes become faulty/unsafe, some channels in V IN 2 around the faulty/unsafe nodes are used as the detours of those channels in V IN 1 passing through the faulty/unsafe nodes, i.e., the adaptability in V IN 2 is transformed to support fault-tolerant deadlock-free routing. Using information on the state of each node's neighbors, we have developed an adaptive fault-tolerant deadlock-free routing scheme for n-dimensional meshes and hypercubes with only two virtual channels. In an n-dimensional hypercube, any pattern of faulty nodes can be tolerated as long as the number of faulty nodes is no more than dn=2e. The maximum number of faulty nodes that can be tolerated is 2 n?1 , which occurs when all faulty nodes can be encompassed in an (n ? 1)-cube. In an n-dimensional mesh, we use a more general fault model, called a disconnected rectangular block. Any arbitrary pattern of faulty nodes can be modeled as a rectangular block after nding both unsafe and disabled nodes (which are then treated as faulty nodes). This concept can also be applied to k-ary n-cubes with four virtual channels, two in V IN 1 and the other two in V IN 2. Finally, we present simulation results for both hypercubes and 2-dimensional meshes by using a variety of workloads and fault patterns. The opinions, ndings, and recommendations expressed in this paper are those of the authors, and do not necessarily reeect the views of the funding agencies.