Adaptive Routing Techniques for Distributed Communications Systems

The concept of a netted or distributed communications system is significant because it offers a major advantage over conventional communications systems, i.e., a better chance of providing surviving lines of communication after an attack on the system. During and after an attack, however, an effective adaptive routing technique is necessary to adjust the routing tables of the message-switching control system to the changing situation. Previously investigated adaptive routing techniques are shown to be insufficient for the task, and a number of promising alternatives are formulated and investigated such as some stochastic techniques, which use information on messages passing through the network to adjust the tables, and some deterministic techniques, which use dynamic programming or graph-theoretic algorithms to recalculate changes in the tables from observed changes in the network. Each alternative has operational advantages, which make it good for certain types of communications systems, and certain disadvantages if applied to others. Regions of applicability and inapplicability, with respect to the above system parameters, are given for the various techniques. Recommended techniques are specified for two currently proposed applications of distributed communications systems. A number of research areas are pointed out in which further efforts in the analysis and simulation of adaptive routing techniques will have useful payoffs in the design of future distributed communications systems.