Research On Key Technologies Of Load Balancing For ID/Locator Separation Networks

Recently, among all kinds of future network architectures proposed in the Internet research community, Identifier (ID)/Locator separation based network architecture which is one of the main directions of the network development has the advantage of scalability, mobility, multihoming, security and etc. However, regarding to the ID/Locator separation networks, the research on the network performance, network re-source utilizing efficiency and control overhead still does not get enough improvement even though they are very significant to the network scalability and deployment, the stability of network application and quality of services of users. Therefore, in this thesis, based on the ID/Locator separation networks, network load balancing and relative topics are studied. The main contributions of the thesis are as follows:1. The current traffic engineering approach based on the ID/locator separation networks cannot achieve load balance well in a realtime, distributed and low control overhead way simultaneously. For this issue, it is proposed a realtime and distributed load balancing mechanism based on the ID/locator separation networks in the thesis. In the mechanism, the router which performs path selection need neither link status, traffic demand nor the optimal route provided by central controller. Meanwhile, the mechan-ism raises neither the problem of convergence nor that of stability. The mechanism only brings very little control overhead. To evaluate, the analytic formula of the performance is defined in the thesis. The mechanism and the packet schedule algorithm are imple-mented in simulator. The unused link ratio, normalized link utilization, traffic ratio and control overhead are compared and analyzed in the thesis. The results show that, the proposed mechanism effectively improves the network resource utilizing efficiency and the network performance while only brings very little control overhead.2. For the purpose of improving network resource utilizing efficiency and perfor-mance in a reliable way, in the load balancing mechanism based on the ID/locator sepa-ration netowks, a load balancing approach is proposed based on the feedback principle. Load balance of the whole network is achieved more effectively by establishing the close-loop control model based on the classical control theory and then by designing the stream distributing algorithm. This approach derives the ideal output rate of each router interface by solving the close-loop control model based on the knowledge of network status measurement, empirical estimation of network performance and performance re- quirement and distributes each stream by using the proposed algorithm based on the so-lution of the model. The approach keeps all merits brought by the load balancing me-chanism in the ID/Locator separation networks, and introduces no additional control overhead. To evaluate the performance of the approach, a simulation is performed. The results show that, comparing to the algorithm proposed in the section2and the classical scheduling algorithm, the approach reduces the network maximum link utilization ef-fectively, and improves the network performance and the network resource utilizting efficiency.3. In the ID/Locator separation networks, an optimized routing protocol based on flat identifier is designed to achieve availability and integrity of the load balancing me-chanism, routing scalability and network configuration complexity reduction. Through modifying the message format and internal data structure of the current routing protocol, the proposed protocol introduces the automatically generated flat identifier to be used as the locator in the ID/Locator separation networks, and takes the flat identifier to be used for packet forwarding. The protocol reduces the scale of the intradomain routing table sharply and simplifies the network configuration evidently. Based on the real topologi-cal data gained from network operators, the analytical formulas of network performance to be evaluated are given. The results show that, the protocol improves the routing sca-lability, and reduces the network configuration complexity and the control overhead ef-fectively.4. In general case, optimizing the network power consumption always stands for reducing the available network resources, while load balancing totally asks for max-imizing the network resource utilization as much as possible. Therefore, for solving the contradiction and achieving the duple target of energy efficiency and load balance, an approach which can simultaneously optimize the network power consumption and achieve load balance is proposed. This approach contains a trade-off model which is nonlinear programming and a heuristic algorithm. The objective function of the model contains the link cost brought by both power consumption and traffic load. Through in-troducing the power-bandwidth formula and designing the cost functions, the collabora-tive optimization of link power consumption and link traffic is performed. Meanwhile, the near-optimal solution to the optimization can be quickly derived by using the pro-posed heuristic algorithm. To evaluate the effect of the approach, the local optimal solu-tions are derived by importing the traffic demands and the real topologies into the pro-gramming model. The solutions of the model are compared with the results from the proposed heuristic algorithm and other models. The results show that, the model and the heuristic algorithm effectively optimize the network power consumption and achieve network load balance. Meanwhile, comparing to the trade-off model, the heuristic algo-rithm owns much less time overhead.