Date of Award

Spring 2004

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Computing Sciences - (Ph.D.)

Department

Computer Science

First Advisor

Nirwan Ansari

Second Advisor

Teunis J. Ott

Third Advisor

Edwin Hou

Fourth Advisor

Joseph Y-T. Leung

Fifth Advisor

James A. McHugh

Abstract

The IP over Wavelength Division Multiplexing (WDM) network is a natural evolution ushered in by the phenomenal advances in networking technologies and technical breakthroughs in optical communications, fueled by the increasing demand in the reduction of operation costs and the network management complexity. The unprecedented bandwidth provisioning capability and the multi-service supportability of the WDM technology, in synergy with the data-oriented internetworking mechanisms, facilitates a common shared infrastructure for the Next Generation Internet (NGJ).

While NGI targets to perform packet processing directly on the optical transport layer, a smooth evolution is critical to success. Intense research has been conducted to design the new generation optical networks that retain the advantages of packet-oriented transport prototypes while rendering elastic network resource utilization and graded levels of service.

This dissertation is focused on the control architecture, enabling technologies, and performance analysis of the WDM burst-switched long haul and Metropolitan Area Networks (MANs). Theoretical analysis and simulation results are reported to demonstrate the system performance and efficiency of proposed algorithms.

A novel transmission mechanism, namely, the Forward Resource Reservation (ERR) mechanism, is proposed to reduce the end-to-end delay for an Optical Burst Switching (OBS)-based IP over WDM system. The ERR scheme adopts a Linear Predictive Filter and an aggressive reservation strategy for data burst length prediction and resource reservation, respectively, and is extended to facilitate Quality of Service (QoS) differentiation at network edges. The ERR scheme improves the real-time communication services for applications with time constraints without deleterious system costs.

The aggressive strategy for channel holding time reservations is proposed. Specifically, two algorithms, the success probability-driven (SPD) and the bandwidth usage-driven (BUD) ones, are proposed for resource reservations in the FRRenabled scheme. These algorithms render explicit control on the latency reduction improvement and bandwidth usage efficiency, respectively, both of which are important figures of performance metrics.

The optimization issue for the FRR-enabled system is studied based on two disciplines - addressing the static and dynamic models targeting different desired objectives (in terms of algorithm efficiency and system performance), and developing a ''crank back'' based signaling mechanism to provide bandwidth usage efficiency. The proposed mechanisms enable the network nodes to make intelligent usage of the bandwidth resources.

In addition, a new control architecture with enhanced address resolution protocol (E-ARP), burst-based transmission, and hop-based wavelength allocation is proposed for Ethernet-supported IP over WDM MANs. It is verified, via theoretical analysis and simulation results, that the E-ARP significantly reduces the call setup latency and the transmission requirements associated with the address probing procedures; the burst-based transport mechanism improves the network throughput and resource utilization; and the hop-based wavelength allocation algorithm provides bandwidth multiplexing with fairness and high scalability. The enhancement of the Ethernet services, in tandem with the innovative mechanisms in the WDM domain, facilitates a flexible and efficient integration, thus making the new generation optical MAN optimized for the scalable, survivable, and IP-dominated network at gigabit speed possible.

Share

COinS