Fault-tolerant Real-time Communication with Reduced Resource Overhead

Providing guaranteed timely delivery of real-time messages in point-to-point networks requires the reservation of network resources such as buffer space and bandwidth based on worst case traffic characteristics (e.g. maximum source rate, maximum transmission delay). For variable-rate sources this causes the network to become underused while at the same time refusing new requests for more realtime traffic. If fault-tolerance of a real-time connection is required, then additional backup resources must be reserved. Current fault-tolerant real-time communication protocols require an overhead that may be 100% or higher of the resources required by the primary channel, even for the simple case of tolerating single node/link failures. Such protocols can be improved by optimising the resource reservation to take into account any failure assumptions for the network and/or dependencies between the routing resources required by different channels. In general, this will reduce the overall resource..