Methods and elements of graph theory and fuzzy logic for communication network management

Abstract

This thesis aims to study the applicability of graph theory and fuzzy logic elements to various management tasks of a communication network. We focus our work on two major fields: communication networks and fuzzy logic with their two fundamental aspects (fuzzy inference and arithmetic properties of fuzzy numbers). Also, every addressed problem is defined and treated based on Graph Theory. From this perspective, we include networks associated with graphs in our research, where we interpret each component as graph elements. We intend to describe each problem with a high mathematical abstraction level and specify them in engineering problems. Also, to solve the problems we apply ad hoc algorithm-based techniques based on fuzzy logic.

To support our purpose we develop and propose solutions to three problems: In chapter 3 we face the server node selection according to a goodness index in the server-client path in a Peer to Peer (P2P) Network). We show in the experimentation that the strategy proposed by us produces the best results concerning the download time in a network with obstacles. In chapter 4 we perform an analysis of the efficiency of different cost functions in high-capacity network links to optimize the traffic load between two nodes. In general, the fuzzy strategies produce the best results over their crisp versions. In particular, a new fuzzy strategy, proposed by us, provides the best results, significantly exceeding the performance of the remaining strategies. In chapter 5 we search for the shortest pair of edge-disjoint paths using fuzzy costs in a high-performance network with priority traffic. We propose an algorithm that always gives a more efficient solution (in quality and stability) when used by nodes with priority communication source in a network with two types of traffic.