APPLICATION OF ARTIFICIAL-INTELLIGENCE TECHNIQUES TO COMPUTER NETWORK TOPOLOGY DESIGN

The topological design of distributed computer networks has been widely studied during the last three decades. Considered for all practical purposes as a very difficult optimization problem with a high risk of combinatorial explosion, it can only be solved by heuristic methods which aim at reducing the search of candidate topologies and provide suboptimal solutions. Taking into account the heuristic nature of conventional design procedures, many researchers have turned to AI techniques for solving some aspects of this problem. This paper presents a new approach based on the artificial intelligence paradigm to deal with the topological design of such computer networks. Its original contribution precisely consists in applying some effective perturbations to a good starting topology in order to reduce total link cost and/or to improve the average transmission delay.