Research On Key Techniques Of Locator/Identifier Separation Network

With the rapid development of Internet, the existing Internet architecture has been confronting many serious challenges such as routing scalability, mobility support, multi-homing and traffic engineering support. It has been widly recognized by both industry and academia today that how to design the architecture for future Internet is one of the hot research topics.To address several problems during the research process of Locator/Identifier separation network (LISN), we study the current key technologies and the application deployment, and propose a LISN model which supports the routing scalability, mobility, multihoming and traffic engineering, and a scalable Identifier/Locator mapping system. And then we focus on the following key technologies: the parallel request algorithm based on prefix hash tree, mapping request prediction algorithm based on generalized regression neural network, Identifier-based routing method for area networks and mobility support method for LISN. We also design and implement a prototype system to validate our work. The major contributions of this thesis are as following:1. Based on the analysis of the existing proposals, a routing scalability, mobility, multihoming and traffic engineering supporting Locator/Identifier separation network model (RSMT-LISN) is proposed. RSMT-LISN is designed from six aspects such as terminologies, phisical structure, logical structure, protocol structure, data structure and working principle. RSMT-LISN can support both IPv4 and IPv6, and have good scalability. Besides, RSMT-LISN can also support mobility, multihoming and traffic engineering.2. Considering the advantages and disadvantages of the mapping systems based on DHT, a scalable Identifier/Locator mapping system (SILMS) based on the core-edge separation idea is proposed. SILMS has a two-level hierarchical structure which is composed of the bottom-level mapping system (BMPS) and the top-level DHT-based mapping system (TMPS). If the communicating nodes are in the same area network, the mapping request messages will be isolated in the area network, and the related mapping request process will be done with the aid of BMPS. In addition, TMPS stores the mapping data which consist of the area network identifiers (AIDs) and the global locators (GLocs) of the corresponding border routers with mapping function, and responses for the mapping request messages from different area networks. The two-level hierarchical structure can efficiently reduce the load of nodes in SILMS and improve the system scalability. Moreover, this structure is good for the large scale deployment of SILMS. Besides, the request latency can be reduced greatly.3. According to the single exact matching problems and common requirements of the existing mapping request algorithms, a parallel request algorithm (PS-Chord) based on prefix hash tree (PHT) is proposed in TMPS. PS-Chord utilizes the arrival distribution feature of mapping request messages, and reduces the mapping request latency. Using the theory analysis and the simulation based on real traffic, the accuracy and feasibility of PS-Chord are proved. Experimental results indicate that PS-Chord is significantly better than the same kind of solutions in the aspects of mapping-request latency, mapping-request hop-count and node load.4. Aiming at the mapping request prediction problem, the basic idea of our proposal is as follows: according to the historical mapping request messages in SILMS, the mapping request prediction is done based on neural network. Thus, the prediction results can give some hints to the mapping data updating processes in the caches of border routers, then the mapping request latency can be reduced where applicable. Based on the above analysis, a mapping request prediction algorithm based on generalized regression neural network (MRP_GRNN) is proposed. MRP_GRNN constructs a mapping request prediction model with the aid of GRNN, selects network parameters adaptively, and updates the prediction model dynamically with the arrival of new data. MRP_GRNN is fast, accuracy, and has superiorities in approximation ability, classification ability and learning speed over Back-Propagation Network or Radial Basis Function Network. In hence, MRP-GRNN can be used to predict the trend of the coming mapping request, and provide the reference data for the mapping data updating in the caches of border routers.5. In LISN, a scalable Identifier based routing architecture (SIRA) is proposed. In SIRA, we present the topology and routing information representation of LISN, and the aggregation method of the routing information. On the basis of SIRA, an identifier based routing protocol (IDRP) is proposed for area networks. In IDRP, we present the neighbor trust relationship contruction, routing information learning, routing table maintenance, optimal route selection and routing information distribution. Then, the packet forwarding principles, multihoming and multipath supporting schemes combined with SILMS are introduced.6. Aiming at the mobility support problem, a seamless mobility support protocol (SMOS) is proposed in LISN. The contributions of SMOS are summarized as follows: 1) SMOS provides different location update schemes for inter-ARN mobility (macro mobility) and intra-ARN (micro mobility) mobility respectively. 2) SMOS addresses the packet loss problem by introducing short term indirections to the routing mechanism realized by using the home agents (HA) as the indirect agents for mobile node (MN) and corresponding node (CN) before the complement of the updating process in the border router of CN's home network. 3) SMOS introduces the shortest path between MN and CN to minimize the packet transmission latency. 4) In the area networks, SMOS uses the entire storage mode to reduce the signal overhead while MN moves inside the area network. 5) Dynamic mapping updating algorithm (DMU). According to the state of MN, the border router of MN's home network deals with the corresponding old mapping data properly. During a period, the old mapping data and new mapping data may coexist in the mapping system. Thus, SMOS can solve the packet loss problem to some extent.7. To validate these key technologies described upon, a LISN prototype system (IdComm) is designed and implemented. We specify the design details of key equipments in IdComm. Finally, a testing environment is built. According to the proposed testing content, the testing results are shown.To sum up, our research is a beneficial exploration of Locator/Identifier Separation network. It has the good theoretical and practical value to the future Internet architecture research. The research has been integrated into the national science and technology support program of China, the natural science foundation of China and our actual project.