Kung-meng Lo

Date of Award

1-1-2008

Document Type

Thesis - ECU Access Only

Publisher

Edith Cowan University

Degree Name

Doctor of Philosophy

School

School of Engineering

Faculty

Faculty of Computing, Health and Science

First Supervisor

Professor Daryoush Habibi

Abstract

The advents of Wavelength Division Multiplexing (WDM) and then Dense WDM (DWDM) allow today's optical networks to simultaneously serve huge amount of traffic requirements. Particularly, the explosion of video, TV and other high bandwidth realtime traffic requires more and more bandwidth that except WDM or DWDM networks, other electronic-based transport networks cannot accomplish. WDM networks have been de factor as the long haul backbone networks around the world. To maintain these facilities is important because with up to 100 Tbps of data flowing through a single fibre in WDM networks, failure can result in huge transmission data losses, breakdown of many service operations and business revenues losses. Designing efficient Routing and Wavelength Assignment (RWA) strategy and protection scheme to optimise capacity efficiency are still open questions. The design efficiency in not only the physical topology of WDM networks but also in the logical topology is becoming very crucial. In this thesis, we study the problem of routing and protection in WDM networks and aim to optimize the amount of bandwidth used to satisfy the requirements of traffic connections between end-nodes.

One significant factor in a WDM network, different from other electronic-based networks, is that it requires assigning wavelength channels along working and protection routes of connections. The placement of full wavelength converters (WCs) at network nodes would reduce the complexity of the problem but significantly increase the investment cost. In contrast, when there are no WCs in the network, the wavelength continuity is required along working and backup routes. This increases not only the complexity of problem but also the blocking probability in the networks. Our objective is to maintain the blocking probability as low as possible and minimise the investment required by placing WCs only at a minimum number of nodes. This problem is referred to as Sparse Wavelength Converters Placement (SWCP) problem.

Considering the SWCP problem with both routing and protection at the same time to maximize the performance of available resources is very important in WDM networks. The design of wavelength routing, protection procedures themselves and the interaction between these procedures allows us to minimize the blocking probability. V./e study the effect of nodal-degrees on the distribution of the traffic load and the blocking probability over the network. From that, we determine the optimal locations to place wavelength converters which minimize the blocking probability. We develop an adaptive RWA approach for wavelength routing strategy. Our results show that our SWCP algorithm selects WCs with the conversion density 0.2 in the entire network and the blocking performance using the adaptive RWA algorithms is very similar to that of full deployed WCs in the networks. We also investigate a protection strategy known as Preconfigured Protection Cycle (p-Cycle). The reason of considering p-cycle as a key protection strategy is that it provides fast restoration for optical connections (lightpaths) clue to its ring-like structure and pre-connected back up paths before failure, as well as its ability to achieve low capacity redundancy because each straddling link on a cycle can provide double protection capacity clue to the nature of cycle. Although current research shows that p-cycles can achieve low capacity redundancy (about 35%), this result is obtained for static traffic and ignores the wavelength assignment problem. Therefore, finding a good set of p-cycles to minimise capacity redundancy in real dynamic networks are the main issues of this study. For the problem of finding good set of p-cycles, we develop three approaches of cycle selection to deliver fast computation with minimum spare capacity. Two approaches are designed for selecting a limited number of qualified candidates, then we use the Integer Linear Programming (ILP) model for determining the optimal set of p-cycles. The third approach is a heuristic approach. To achieve the optimal solution, p-cycle is iteratively selected one by one with the best efficiency until all working traffic is full protected, then a refined cycles selection is performed to complete the selection.

Finally, for the efficiency of p-cycle in real dynamic networks, we analyse the variations of blocking performance within sparse wavelength conversion networks. We also compare the capacity efficiency and computational complexity when using p-cycles as span protection and path protection. These studies are critical and they have not been investigated before.

Download

Share

 
COinS