Engineering a new algorithm for distributed shortest paths on dynamic networks

International audienceWe study the problem of dynamically updatingall-pairs shortest paths in a distributed network while edge update operations occur to the network. We consider the practical case of a dynamic network in which an edge update can occur while one or more other edge updates are under processing. A node of the network might be affected by a subset of these changes, thus being involved in the concurrent executions related to such changes. In this paper, we provide a new algorithm for this problem, and experimentally compare its performance with respect to those of the most popular solutions in the literature: the classical distributed Bellman-Ford method, which is still used in real network and implemented in the RIP protocol, and DUAL, the Diffuse Update ALgorithm, which is part of CISCO's widely used EIGRP protocol. As input to the algorithms, we used both real-world and artificial instances of the problem. The experiments performed show that the space occupancy per node required by the new algorithm is smaller than that required by both Bellman-Ford and DUAL. In terms of messages, the new algorithm outperforms both Bellman-Ford and DUAL on the real-world topologies, while on artificial instances, the new algorithm sends a number of messages that is more than that of DUAL and much smaller than that of Bellman-Ford