Spectrum Flexible Resource Allocation And Optimization Of Key Technology Research In The All-optical Networks

With the rapid development of Internet as well as the population of boardband service, the network bandwidth demand is booming in a high rate. As a result, the future all-optical network evolutes towards an ultra-high capacity and flexibility one. The traditional wavelength routed all optical networks, allocating and managing network resource in an one-size-fits-all manner, lead to inefficient resource utilization, low flexibility and high power consumption. By introducing the bandwidth-varible transmitting and swithing technologies, the flexible bandwidth optical networks utilize spectrum resources efficiently, thus having a wide application prospects. However, owing to supporting finer granularity segementation and dynamic regulation, the flexible optical architecture brings additional challenges on the networking level. Among these, spectra resource modeling, routing and spectrum allocation and spectrum defragmentation are some of the most important issues. To solve these problems, with the support of National863project, this paper mainly focuses on the key technologies in bandwidth flexible optical networks, especially on the problem of resource dynamic allocation and optimization. The main innovative contributions are listed as follows.Firstly, according to the spectrum complex status and multi-constraints in flexible optical network, two spectrum description methods are designed based on the analysis of spectrum utilization. They are spectrum consecutiveness based model and path connectivity based model respectively. In the spectrum consecutiveness based model, we introduce the notion of link spectrum consecutiveness and path spectrum consecutiveness, which denote the available degree of free spectrum on a link and along a path respectively. In the path connectivity based model, a notion of Node Spectral-X eigenvectors is introduced to denote the switching ability for all FSs according to Node spectrum usage Matrix. In addition, a metric of Path Connectivity is defined to represent the maximum variation of node switching ability along the path in flexible bandwidth networks. These two methods provide a theoretical basis on the following routing computation, spectrum assignment and spectrum defragmentation.Secondly, in spect of resource configuration and optimization, it is of great importance to retain the spectrum consecutiveness and completeness in the flexible bandwidth optical networks. How to regularize the occupied spectrum on the paths or links is fundamental issue. In this paper, aiming to optimizing the path spectrum resource, links spectrum resource along the same path and heaviest payload link resource, three dynamic spectrum allocation algorithms are proposed. They are Maximize Path Spectrum Consecutiveness algorithm (MPSC), Maximize Total Link Spectrum Consecutiveness algorithm (MTLSC), and Maximize Heaviest Link Spectrum Consecutiveness algorithm (MHLSC). These three algorithms can reduce the spectrum fragments by maximizing the path or link spectrum consecutiveness during the process of resource allocation. On one side, in these three algorithms, the spectrum resource retains consecutive as much as possible, thus reduce the fragments. On the other side, these algorithms utilize the current fragments on a link or along a path at most, therefore further promoting the network performance. These proposed algorithms are realized on the simulation platform of flexible bandwidth optical networks. Simulation results show that in case of mix-line-rate scenario, the proposed algorithms achieve lower blocking probability and higher adaptivity. Among these three algorithms, MTLSC yields better performance.Thirdly, in the respect of optimized routing algorithms, complying with the spectrum continuity and adjacency constraints, a novel dynamic routing algorithm is proposed and realized by means of Ant Colony Optimization. In our algorithm, the ant-based agents gather path length information and spectrum consecutives information as their private data. Then they continuously read and modify the routing tables according to its prvite data. As a result, the routing table reflects the the current network status. An advantage of our algorithm is that it incorporates the path length and spectrum consecutiveness information into the routing decisions. In this way, the shorter path with less spectral fragments is apt to be selected, thus reducing the blocking probability. Simulation results show that the ACO based routing algorithm reduce the blocking probability by half comparing to the wavelength routing algorithm in the WDM network. In addition, the proposed ACO based routing algorithm achieves high adaptivity in case of mix-line-rate scenario.Lastly, in the respect of spectrum defragmentation, by analyzing the fragments generation reason and its elimination method, adopting blocking sevice oriented trigger mechanism and networking performance oriented trigger mechanism, two defragmentation algorithms are proposed on the basis of path connectivity. They are Maximize Path Connectivity algorithm (MPC) and Path Connectivity Triggered algorithm (PCT). Accordingly, we define a cost-performance-ratio based profitability model to denote the prons and cons of spectrum defragmentation. Simulation results show that compared to the non-defragmentation condition, both MPC and PCT defragmentation algorithms can better improve the performance of blocking probability. In case of light traffic load condition, MPC gains higher defragmentation profitability while in the heavy load condition, PCT is better in spect of defragmentation profitability.