Fault-Tolerant Multicast Communication for Multicomputers.
Rajendra V. Boppana, Suresh Chalasani · 1995
. We describe fault-tolerant routing of multicast messages in mesh-based wormhole-switched multicomputers. With the proposed techniques, multiple convex faults can be tolerated. The fault information is kept locally--- each fault-free processor needs to know the status of the links incident on it only. Furthermore, the proposed techniques are deadlock- and livelock-free and guarantee delivery of messages. In particular, we show that the previously proposed column-path and Hamilton-path based algorithms can be made tolerant to multiple faults using two or three virtual channels per physical channel. Keywords: block faults, fault-tolerant routing, Hamilton path routing, multicast routing, wormhole routing. 1 Introduction Many commercially available parallel computers use mesh or grid based networks for interprocessor communication with a processor and router module at each node [5, 10]. The interprocessor communication functions in a multicomputer are usually handled by a router which...