Anycasting and Multicasting in Adversarial Systems: Routing and Admission Control

In this paper we consider the problem of routing packets in dynamically changing networks, concentratingon two different modes: anycasting and multicasting. In anycasting, a packet has a set of destinations butonly has to reach any one of them, whereas in multicasting, a packet has a set of destinations and has toreach all of them. Both communication modes have a tremendous number of applications, but so far notmuch is known from a theoretical point of view about how to efficiently support these communication modeseven in static networks. We demonstrate in this paper that even if both the network and the packet injectionsare under adversarial control, efficient distributed routing strategies can be found for both modes. Thiseven holds if the adversary is not bounded in the number of injections, and therefore packets may have tobe dropped from overfull buffers. In order to study the performance of our protocols, we use competitiveanalysis. The performance is measured in terms of communication throughput (i.e. the number of successfulpacket deliveries) and space overhead. Our results demonstrate that, in principle, efficient communication ispossible even in such highly dynamic networks as mobile ad hoc networks and peer-to-peer networks. Supported by DARPA grant F306020020550 “A Cost Benefit Approach to Fault Tolerant Communication” and DARPA grantF30602000-2-0526 “High Performance, Robust and Secure Group Communication for Dynamic Coalitions”.Supported in part by the DFG-Sonderforschungsbereich 376 “Massive Parallelität: Algorithmen, Entwurfsmethoden, Anwendun-gen”. Part of the research was done while visiting the Johns Hopkins University, supported by a scholarship from the German Aca-demic Exchange Service (DAAD Doktorandenstipendium im Rahmen des gemeinsamen Hochschulsonderprogramms III von Bund undLändern).