Research On Quality-of-Service And Fault Detection Scheme For Optical Burst Switching Networks

With the increase of communication capacity and modern society's dependence on communication, people declare higher requirements to backbone networks on the bandwidth requirement and Quality of Service (QoS) and thus speed up the research of optical transport networks. At present, three switching paradigms have been proposed to realize all optical transport; there are Optical Circuit Switching (OCS), Optical Packet Switching (OPS) and optical burst switching (OBS), in which, OBS, which combines the merits of OCS and OPS while avoids their disadvantages, is regarded as the most competitive transmission and switching technology for next generation all optical networks.Optical burst switching networks can carry diverse traffic coming from higher layer. How to meet the Quality of Service(QoS) requirement of such traffics and how to implement fault management in OBS networks, will attract more researcher's attention at present as well in future, it is also the subject of my work. And this dissertation focus on the study of the factors which affect the QoS of OBS networks including data channel scheduling algorithms operated at core node and contention resolution scheme in core networks as well as fault monitor mechanism in OBS networks.It can be found that traditional data channel scheduling algorithms as LAUC, has a low complexity, but it also bring high loss rate of data burst; Another algorithm, LAUC-VF can use any void between random two data-bursts which can improve data burst loss but also cost a long run time. Under this background, two scheduling algorithms named Data-Burst-Length-Based (DBLB) and Void-Based (VB) are proposed in chaper 2, which combines the merits of the LACU and LAUC-VF and simultaneously avoids their shortcomings. The main ideas are: set a threshold based on the average length of the received data burst in a time windows T, or according to the void conditions in data channels, if the incoming date-burst's length is longer than the threshold, using LAUC to schedule this burst. Otherwise, using LAUC-VF to schedule them. Simulation result show that the two algorithms can effectively reduce data burst loss while just cost a short run time. In chapter 3, deflection routing is discussed which is an effective contention resolution for Optical Burst Switching network. To prevent the loss rate ascending of normal (non-deflected) bursts appeared on deflected routing, two deflection routing schemes based on contention-control with and without priority difference are proposed individually. By using defined check function, the proposed schemes make some operations to guarantee the QoS performance of non-deflected burst, which include deflecting conditionally or dropping the contention burst, as well as limit the deflected high priority burst's ability to preempt reserve resource of non-deflected burst. Simulation results show that these schemes can efficiently prevent deflected burst competing with non-deflected burst on deflection route, and obviously improve the burst loss performance of entire networks.Deflection routing has been research from the node scope in chapter 3, which belongs to local contention resolution scheme. In chapter 4, the contention resolution problem is investigated from the network scope. Refer to the routing schemes in OBS networks, shortest path first routing approach is frequently used, which can lead to high blocking probabilities and poor adaptability in different network topology, at the same time it can not reflect the real-time change of traffic flow. So a load-balancing routing mechanism in OBS networks is proposed in this dissertation, which chose a path that can balances the assigned wavelengths of every links. Simulations demonstrate that this method is better than shortest path first routing in terms of block probability, and has similar performance in transmission delay with shortest path first routing, furthermore, It have good ability to reflect the traffic flow variation in real time and can be adaptive to diverse network topologies.Chapter 5 continues on the study of load-balance routing methods discussed above. The proposed routing method here, not only ensures the whole network balance, but also considering the following problems: the network performance assurance of resisting the destruction, the reduction of cost as well as the implement complexity. Accordingly, two types of protection strategies, shared-path protection and shared-link protection, are discussed, the optimization objective here is to minimize the total needed wavelengths in the entire network, and NSFNET is used as an example. By solving ILP models for multi-fiber OBS networks, the features of different protection strategies are analyzed. The influence of wavelength conversion on the optimization objective is also investigated. The simulation results show that this method can effectively reduce the number of overall wavelength in need as well as the cost of the network.Based on the previous study of network architecture and signal protocol, a new fault monitoring mechanism for Optical Burst Switching Networks is proposed in chapter 6. The main objective of this mechanism is to improve the traditional optical network monitoring methods, which will generate abundant false alarms in OBS circumstance. By using data channel information contained in burst header packet of OBS, the mechanism can control the alarm-permitting time and discard redundant false alarming information. So it can provide reliable alarming information for fault location module, and effectively decrease the number of false alarms while almost no missing alarms happen. Simulation results validate that, the shorter time the fault detection needs, the more false alarms are filtered.Based on the fully studies of the property of OBS networks and technologies of optical performance monitoring, a concept of probe burst is introduced in the last chapter, accordingly, a novel performance and monitoring scheme of data channel for optical burst switching networks is investigated, it works as follows: By sending probe bursts on a wavelength channel periodically or in a certain way predefined by the protocol, and detecting their abnormal status in the next-hop node, channel state between two adjacent nodes can be acquired. Therefore, It can be monitored in real time way when a fault or failure has occurred, moreover, while the proposed scheme is applied, both hard failures that interrupt or deteriorate data channels in unexpected ways, or soft failures which caused by the progressively degradation of transmission quality, can be effectively detected in a short time, meanwhile it is of practicability and maneuverability. But the method base on per-hop detection is not practical due to its high cost, so an economical fault detection method, employing cycle cover theory, is proposed in terms of costly sing-hop test module in meshed optical burst switching (OBS) networks. Three kinds of cycle cover finding algorithms, heuristic depth first searching (HDFS), shortest path Eulerian matching (SPEM) and Heuristic spanning-tree (HST) based cycle finding, are used to find cycle cover for meshed networks. Then each cycle is assigned a probe module, and a fault detection mechanism based on probe cycle cover is formed. We show through computation and statistic that the proposed mechanism can reduce the cost of fault detection deeply compared to single-hop test scheme in meshed OBS networks.