Research On Traffic Grooming Algorithms In Optical WDM Mesh Networks

The rapid development of Internet drives the research of optical network dramatically. With the explosive increase in network traffic and the emergence of high performance optical network devices, such as optical cross-connect (OXC) and optical add/drop multiplexer (OADM), wavelength-division multiplexing (WDM) technology have emerged as a promising candidate for next-generation networks. Since low-rate traffic services (e.g. IP services) are widely in use, optical networks are required to rapidly provide high-quality and multi-granularity bandwidth services to users. Intelligent optical network (ION) with independent control plane is introduced. With the developing of WDM, the processing speed of electronic elements such as IP routers, switchs and multiplexers become the bottleneck of transmission. Since each wavelength can be operated at very high speed, e.g., OC-48, OC-192 and OC-768 (the corresponding rates are 2.5Gb/s, 10Gb/s and 40Gb/s, respectively), the huge bandwidth gap between the capacity of a wavelength and the bandwidth required by low-rate traffic streams, e.g., OC-1, OC-3 and OC-12 (the corresponding rates are 51.84Mb/s, 155.52Mb/s and 622.08Mb/s, respectively), also being deeply dicussed. To accommodate such kind of low-rate traffic streams (or connections) with one lightpath will lead to inefficient resource utilization. At the same time, it is impossible to establish end-to-end lightpaths for all the connection requests, due to the limits of the number of wavelengths per fiber and the number of transceivers per node. So it is necessary to investigate how to efficiently set up connections for these traffic streams. Traffic grooming, which can also be called grooming, routing and wavelength assignment (GRWA), can aggregate low-rate connections onto high-capacity lightpaths to make efficient use of the bandwidth to meet this problem. In this paper, with the control plane deploying generalized multi-protocol label switching (GMPLS), the optimization problem for intelligent optical network is studied, and the focus is traffic grooming problem in WDM mesh networks, which covers the following: the structure of nodes in grooming networks, the construction of virtual topology for traffic grooming, QoS-specified traffic grooming, and the survivability of grooming networks.The structure of nodes in grooming networks is analysed in chapter 2, then a novel method of layered virtual topology constructed for traffic grooming is proposed, which is the basis for the algorithms decribed in the latter chapter. Considering engineer applications, we firstly focus on the sparse multifiber and wavelength assignment, then the traffic grooming with different DXC nodes based on layered virtual topology is discussed. (1) Constrained by the price and technology level, not only the wavelength converters laid in WDM network but the wavelength per fiber (WPF) are limited. But there usually existed some spare fibers in links for over-building, so it is important to make full use of these fibers to lower the blocking probability in WDM networks. We proposed asparse multifiber and wavelength assignment algorithm based on dynamic traffic serviced WDM networks to optimize WPF. By assigning sparse multifiber, the blocking performance can be obtained with fewer number of WPF. (2) Since the OXC with grooming capability is still on lab research, in operating networks, the OXC has only wavelength switching capability and extra aggregation/deaggregation NEs such as DXCs are needed to support low-rate traffic streams switching. Considering the cost of networks construcing, the DXCs must be sparsly laid. Considering the node model of WDM mesh networks, layered virtual topology graph is proposed, and traffic grooming is executed on the virtual topology.Most of the researches related to traffic grooming focused on ring networks, and more concerns are put to WDM mesh netwroks in recent year, but fewer concern has considered the QoS. As we known, delay is the most important issue for QoS, to meet the delay performance of QoS, we must consider: (1) The hops of lightpath, which stands for the delay of data-processing in the intermediate nodes and the fiber transmission; (2) The max load of lightpath, when traffic passes through a heavy-loaded lightpath it may deteriorate the delay of node processing; (3) The connections of larger amount of traffics affect more to the mean delay of network. In chapter 3, we propose a QoS-specified traffic grooming algorithm named BMST considering both the hops of lightpath and load balance in the network. Simulation results show that BMST can improve the network load balance with little change of lightpath hops thus provides better QoS.Since a lightpath may carry lots of dada streams, any failure can lead to significant loss upwards a Terabit of traffic per second. Effective syrvivability mechanisms in WDM grooming networks must be investigated. In chapter 4, we discuss the traffic grooming based on shared protection in two ways: (1) Present a shared path protection traffic grooming algorithm based on wavelength layered-graph. The basic idea can be described as follow: Firstly, set up the link-disjoint work paths (single-hop lightpaths) and their corresponding shared protection paths on the same wavelength plane, then use the single hop work lightpaths to construct virtual topology, low-rate connections are groomed in the virtual topology. The simulation with different connectivity networks shows that, the alogrithm performs better than others. (2) Fortunately, most current practical network such as NSFNET and CERNET are Hamiltonian. In this paper we propose a shared link protection based traffic grooming algorithm through Hamiltonian cycle, to construct virtual topology with the Hamiltonian cycle and the link-disjoint work paths outside the cycle, low-rate connections can be groomed in the virtual topology accordingly.Most of the earlier schemes for survivability just provide 100% guaranteed path protection or no protection according to the connection requests; differentiated reliability services are not taken into consideration. But, providing fault tolerance at an acceptable level in WDM networks has become a critical issue due to the size of the current and future networks and diverse QoS requirements for multimedia services. There exists a fact that the failure of components is probabilistic especially for large networks. In such a probabilistic environment network service...