Research On Topology Optimization In Structured Peer-to-Peer Network

In recent years, P2P technology has become the focus in the internet research area. Compared with other network models, P2P network has decentralized, scalable, robust, high performance, and many other notable features, which makes P2P technology and its application concerned by lots of people. In P2P systems, overlay network is built on the logic concept, without much concern about the underlying physical topology. Therefore, the logical shortest route may not be the physical shortest one. This topology difference between logic and physical layer greatly reduces the efficiency of resource locator. This problem is known as topology mismatch problem.This dissertation carried out in-depth research on the topology and working principle of each generation in the development of P2P network, and summed up the topology mismatch problem in structured P2P network and the inadequacy of the existing solutions. We focused on the working principle of Chord network, and according to the topology mismatch problem in Chord, this dissertation designed a Quasi-Chord model, which can be topology aware, and propose a new routing table structure and some key algorithms.The main issues and contributions of this dissertation are as following:Firstly, this dissertation introduces the network coordinate system into structured P2P network. Simulate the whole Chord network into a two-dimensional geometric space with a complete coordinate system, so that every node has a distinct corrdinate in the space. This method can effectively solve the problem in the traditional landmark clustering algorithm that the neighboring nodes can not be precise positioned.Secondly, an improved structure of the routing table is proposed. The improved table includes a clockwise table and a counter-clockwise table. After comparing the key of target resources with the identifier of the node itself, the model can choose which table it should look up into. The new table structure can greatly improve the efficiency of resource locator.Thirdly, this dissertation designs a complete topology-aware model Quasi-Chord. The model is based on the ideaology of Chord, but sets up the model based on the underlying physical network topology. So the Quasi-Chord model can effectively solve the logical network and the physical network topology mismatch problem.Finally, this dissertation puts forward a new routing algorithm, and nodes joining and leaving network maintenance algorithm. In reality, the nodes joining and leaving network is random, and undisciplined. These algorithms can guarantee our model will not collapse in the fluctuation of the network and resume the correct routing information as soon as possible.Simulation experiment shows that Quasi-Chord model can reflect the underlying physical topology, improve the efficency of resource locator and reduce the network traffic.