Resource allocation for quality of service based routing in computer networks

Waseem A Sheikh, Purdue University

Abstract

This research considers the challenges involved in resource allocation for quality-of-service (QoS) routing in computer networks. We focus on four main challenges: QoS multicast routing with resource allocation, joint optimized QoS routing with resource allocation, QoS routing with resource allocation in heterogeneous networks and a resource allocation scheme for wireless networks. Our underlying network model captures the functional relationship between QoS parameters and available resources. Our framework also incorporates the relationship between various QoS parameters. We address the QoS multicast routing with resource allocation problem by formulating it based on our network model. We present an integer program (IP) formulation and a polynomial time heuristic for the QoS multicast routing problem. Using simulations we show the superior performance of the proposed heuristic. In the context of the second challenge, we formulate a QoS routing problem which maximizes reliability and minimizes jitter delay. We propose a dynamic programming based algorithm which finds paths with maximum reliability and minimum jitter delay. Using simulations we provide insight into the tradeoffs involved between maximizing reliability and minimizing jitter delay. Under the third challenge, we formulate and solve QoS inter-domain routing with resource allocation problem for heterogeneous networks employing different queuing service disciplines. Our routing protocol finds optimal inter-domain paths with maximum reliability while satisfying the end-to-end jitter and bandwidth constraints. Finally we address the problem of bandwidth allocation and data droppage in wireless networks. We present an optimal proportional bandwidth allocation and data droppage scheme to provide differentiated services (Diff-Serv) for streaming multimedia data in wireless networks. We demonstrate through simulations that our scheme exhibits the desirable features of absolute and relative DiffServ approaches.

Degree

Ph.D.

Advisors

Ghafoor, Purdue University.

Subject Area

Electrical engineering

COinS