Multiple Tabu Search Algorithm for Solving the Topology Network Design

Abstract

A communication network can be illustrated by set of nodes (or switches) and links (or arcs) where all nodes are connected by links.The typical communication network structure is composed of two levels.The first one is backbone network and the second one is local access network (LAN).The backbone network dedicate for delivery of information from source to destination (end to end) using its switching nodes.The LAN is typically centralized system that allows users to access hosts or local servers.This chapter focuses only on the backbone network design considered as distributed network.The advent of low cost devices has led to explosive growth in communication networks.The topology network design is a part of network planning which finds a topology in order to satisfy any constraints.One of the major advantages of distributed network over the centralized system is their flexibility to improve system reliability.The reliability of a system depends not only on the reliability of its nodes and links, but also on the topology of network.A completely connected network has the highest network reliability while the simple loop (ring) network has the lowest network reliability.Many researchers have studied the designing of the network when considering system reliability as a constraint or an objective.The reliable network design problems state as source-sink and all-terminal network reliability (also known as overall reliability).The problem of optimal topology network design that selects the links connections that either maximizes reliability or minimizes cost can formulate as a combinatorial problem.This problem is a well known NP-hard problem and very difficult to solve.For such problem, many researchers have studied with enumerative-based and heuristic method.Jan (1993) developed an algorithm using decomposition based on brand and bound to minimize link cost of communication network subject to reliability constraint.Aggarwal et al. (1982) employed greedy heuristic approach to maximize reliability given a cost constraint for networks with different reliability of links and nodes.Pierre et al. (1995) also used simulated annealing to find the optimal design for packet switching networks where considering the delay and capacity, but reliability was not.For the network design, Kumar et al. (1995) developed a genetic algorithm (GA) considering diameter, average distance, and communication network reliability and applied it to four test problems of up to nine nodes.Darren & Smith (1998) presented a GA approach for minimum cost network design problem