Reliable connection oriented networks

Abstract

When protected network connections on a connection oriented network are disrupted because a network resource fails (e.g. the severing of a fiber cable upon which multiple links are multiplexed), they must be rerouted around the network failure. This invention expedites this rerouting and permits it to be carried out in a distributed manner similar to conventional routing. Advantageously, the amount of link bandwidth reserved for rerouting of failed connections is reduced relative to conventional methods. For each protected connection, for each anticipated failure condition, a 'contingent path' is selected shortly after the connection is established. This contingent path is then used to reroute that connection in the event of the occurrence of that failure condition. To the extent that only one of the anticipated for network failures occurs at a time (possibly affecting many links, and all connections using those links), the corresponding set of activated contingent paths will all fit within the reserved bandwidth of each link. This follows from the rules used in the prior selection of the contingent paths whereby link reserved bandwidth is allocated to contingent paths on the basis of anticipated failure conditions. Advantageously, in case of a network failure, most or all protected connections can be re-established on the first try.