The decomposition of a communication network considering traffic demand interrelations

Abstract

It is well-known that the analysis leading to the exact, or near exact, evaluation of the performance of communication networks becomes intractable as the system grows in dimension and complexity. In order to reduce the dimensionally of the analysis of a large-scale data network, a new decomposition algorithm is proposed in this paper. The algorithm yields an aggregation policy (the grouping of nodes and links in different zones of the network). The solution of the proposed new algorithm is aimed at making possible the independent analysis of the so-obtained aggregated ares (subsystems) and then the analysis of the complete network on the basis that the aggregate areas are weakly related to the rest of the network. The new decomposition algorithm takes into account not only the structural information of the network, but also the traffic demand interrelation, and is independent of the routing strategy. Furthermore, under an appropriate re-definition of parameters, the proposed decomposition algorithm can be applied to a number of other related problems. The paper concentrates on the generic development of network decomposition. The results show that the algorithm can be used for a wide variety of network topologies and traffic demand patterns. Four different versions of the algorithm are implemented and these result in a saving of up to 28% in computational time for a network consisting of 188 nodes and 772 origin-destination (O–D) traffic demand pairs. The application of the solution obtained from use of the algorithm will be reported elsewhere.