Reroute strategies for fault tolerant packet switching networks

The design and implementation of reroute strategies for fault-tolerant packet switching networks are discussed. A key facet of the network response to trunk failures is fast and reliable failure detection. A good way to detect trunk failures locally is to have the adjacent nodes count errored frames and flags at the data-link level. Failures can be detected at the network edges by having local nodes transmit signaling packets to each source node affected by the failure. The implementation complexity of a strategy depends on the packet forwarding technique in use. The author discusses efficient methods to effect source and local reroutes for two different packet forwarding techniques in common use today. He examines the engineering cost-performance tradeoffs for reroute strategies. It is shown that the cost of restoration capacity required is considerably less than 100% for a nominal degradation in performance during reroute.<>