Information dissemination in communication networks

Abstract

This thesis is concerned with the information dissemination process in a communication network, whereby a message, originated by one vertex, is transmitted to all vertices of the network. It is assumed that a vertex can either transmit or receive a message and an informed vertex can transmit it to only one of its neighbors at a time. For the case of uniform edge transmission times, a moderate amount of work has been done to date. However, for the nonuniform case, few results have been found in the literature. The main objective of this thesis is to investigate the case of nonuniform edge transmission times. In a tree-type network, it is shown that a simple ordering policy achieves the minimum broadcast time under the minimax criterion, and based on this, an O(NlogN) algorithm is derived to determine the broadcast center. Two extended models are considered. One is the model with an ACK scheme and the other is the model with due dates. Broadcasting under the minimum criterion is also considered. The necessary and sufficient conditions is established for the optimal call sequencing in a tree. It is shown that the optimal call sequence is determined by edge transmission times and the sizes of the corresponding subtrees. An O (NlogN) algorithm is designed to determine the broadcast center in a tree under the criterion. In a general-type network, some heuristics are designed based on the results of the tree-type network and vertex degrees. Through the computer simulation, their performances are evaluated. It is observed that in a sparse that in a sparse network, it is good to transmit a message first to the vertex with the largest number of uniformed neighbors. But it is also observed that in a dense network, one must take account of the eccentricity in determining call sequences.