Studies On Performance Of Time-slotted Optical Burst Switched Networks And Related Key Technologies

Exploring a novel all-optical network to support fast growing Internet services is the important task to develop next generation optical networks, while all-optical switching technology is the key issue involved in such tendency. So far the traditional electronics layer switching technologies, such as SDH/SONET transport network that is based upon O-E-O processing, cannot meet the bandwidths needs of future's multimedia services due to the constrains of Mole's Law, and cannot make fully use of bandwidth advantages that benefits from Dense Wavelength Division Multiplexing (DWDM) either. Because optical component techniques are far from mature, today, developing a novel and applied all-optical switching technology and related networking mechanism is significant for the development of next generation backbone.This dissertation dedicates to investigate key techniques incurred in Time-Slotted Optical Burst Switched Network (TS-OBS) - a potential technology for next generation optical network. Our work aim at setting up its network architecture and control plane, developing related supporting algorithms and protocols, and evaluating network performance especially blocking performance. We hope the work in this thesis will provide reasonable theoretical evidence and contribute to organize experimental work for all optical network technologies deployed in the future's so-called Optcial Internet.The content of this dissertation as well as part of results from research can be concluded as followings:1. We summarize features and history of development for three types of optical switching technologies including optical packet switching, optical Time Division Multiplexing (TDM) and conventional Optical Burst Switching (OBS), and compare advantages and issues between different technologies used in current engineering network in reality and experimental network studied in laboratory. 2. For TS-OBS, we propose a consolidated network architecture, confirm its control plane and operation mode for both edge and core nodes, and time slot channel portioning. Under above network architecture, we analyze the key issues to effect blocking performance of TS-OBS network, i.e., mechanism of Burst Assembly, partitioning methods for time slot channel, routing technologies as well as network restoration strategies.3. It is the first time by us to propose a basic framework to evaluate the performance of overall network blocking performance. The model can not only analyze the blocking performance of TS-OBS, but also extend the proposed basic model to be a differential model that can estimate blocking performance under multiple types of traffic circumstances in the network, which is also confirmed by our simulation experiment. Simulation results indicate that the blocking performance can be improved by increasing number of wavelengths, fibers and slots granularity within a frame. Our related work in this field can be summarized as follows:1) Propose a theoretical framework at the cost of low computational complexity to calculate overall blocking probability for TS-OBS network.2) Choose reasonable model the results derived from studying and comparing a ranges of sub-models including link blocking model, path blocking model, offered load model and multi-fiber model, etc.3) Evaluate the effects of different changes, such as the number of wavelengths, number of fibers and slots within a frame, on the overall average blocking performance in terms of proposed framework. Also, we verify the calculation result by simulation.4. Investigate the time slot assignment problem involved in circuit switching traffics as well scheduling problem involved in best effort traffics, which exist in both core and edge nodes in TS-OBS network. We also develop corresponding algorithm for above problem respectively. Work in this issue can be listed as follows:1) For circuit switching traffics, we give a mathematical definition and propose a revised off-line TS-Greedy time slot assignment algorithm. Moreover, we analyze the differentia between the proposed greedy algorithm and optimized theoretical solution, and the analytical result and simulation result are provided for further discussion.2) For best effort traffic, we convert the original scheduling problem to a multi-dimension Integral Variable sized Bin-packing Problem. We also deduce upper and lower bounds of competitive ratio for three on-line scheduling algorithms. Following our theoretical work above, we prove that Best-Fit and First-fit is prior to Worst-Fit algorithm, and this is confirmed by our further simulation.3) For best effort traffic, we study the overall blocking performance for several on-line scheduling algorithms used by solving Bin-packing problem in reality, especially survey the blocking performance of Best-Fit algorithm by changing a range of network parameters.5. A dynamic load balancing routing technique that is special for best effort traffic in the TS-OBS network is also proposed in Chapter 5. Based upon virtual multi-wavelength topologies based routing algorithm combined with fiber selection algorithm, the proposed routing technique can consider both channel capacity and number of hops in the process of making route decision by changing the value of weight function. With adjusting a range of different weight function, we study the overall network blocking performance by simulation. And Best-fit algorithm is chosen to be a candidate method in later experiment to extensively survey the impacts of scheduling algorithms on packet loss rate in the network environment. Experimental results indicate the proposed routing technique is effective to smooth congestion occurred in the TS-OBS network.