Research On Several Key Technologies Of MP2P In Heterogeneous Network Environment

As a distributed computing model, P2P (Peer-to-Peer) is self-seeking individuals or institutions share network resources and services by direct exchange, which meets the future network development. However, with the wireless access technologies emergence and the enhancement of the mobile terminal function, a large number of distributed applications appear in heterogeneous network environments. Therefore, the research on key technologies of MP2P (Mobile Peer-to-Peer) has high theoretical and practical value in the heterogeneous network environment.A MP2P resource discovery algorithm based on VL-Chord (Vector Location Chord) is presented. The whole network is divided into different regions by the cellular theory and polar coordinates, and the new routing table is established which includes three-dimensional location vector information:large ownership area, distance and direction. The new routing table reflects the cross-layer optimization, then the rapid resources discovery is realized; VL-Chord has bidirectional searching functions, and overcomes the common topology mismatch problem. DCM (Dynamic Cellar Management) mechanism realizes the dynamic allocation of resources, reduces the expenses of the nodes'mobility. Simulations based on OMNeT++ show that VL-Chord algorithm reduces the routing hops and latency with the MP2P network size increasing; meanwhile, enhances the capacity of load balancing. Performance of simulation datas and theoretical analysis values are basically similar, which further reflects the validity and credibility of the proposed algorithm.A Ring Dynamic Adjustable Self-organization resource discovery algorithm (RDAS) for MP2P network is proposed. The MP2P network is divided into concentric ring region, whose radius is in the exponential increase; the node failure rate is defined. RDAS algorithm approximates goals in exponent-level, has high routing efficiency, and RDAS avoids the topology mismatch. RDAS algorithm overcomes the impact of mobility, when the part of the network connection is invalided; the whole failure rate can be dynamically adjusted in according with the node moving speed and ring radius, which reflects better flexibility and fault tolerance. Simulations based on OMNeT++ show that the routing efficiency of RDAS algorithm is high, the failure rate is changed with the node speed and radius parameter; and the dynamic balance of the failure rate can be dynamically adjusted by the two parameters.A MP2P traffic intelligent management model (BP-IM) based on neural network is proposed. The network is modeled by the semi-distributed structure, the BP neural network (BPNN) is to measure the traffic effectively; the assignment of traffic is dynamically adjusted by the MP2P traffic priority table. The BP-IM model has the features of flexible configuration, efficient detection, and easy to be extended, reducing the complexity of the model. Simulations based on OMNeT++ show that: compared to the traditional model, higher-level traffic processing delay is small by the BP-IM model, which saves the bandwidth resources and avoids network congestion. MP2P traffic processing delay decreases with increase of the traffic priority, which reflects the intelligent management concept of the BP-IM model.A master-slave structured trust punishment model (MSSP) based on the MP2P network is proposed. The nodes are classified into the credibility server and ordinary nodes by the hierarchical management, and are given the different data structure; and the credibility server nodes are responsible for the credibility value update, and MSSP is a hybrid type of trust model. Transaction credibility takes into account the region credibility and the recommendation one, making it more resistant to aggressive. Credibility binary punishment mechanism effectively reduces the impact of malicious nodes. Simulations based on OMNeT++ shows MSSP model have good performance both in static network environment and in dynamic one; especially in the dynamic MP2P environment, transaction success rate of MSSP model has an increase of 22%; the MSSP characteristic of fast convergence is also consistent with the requirements of MP2P trust model.On the basis of the above, super-node selection and network QoS guarantee are also researched, and presents relevant algorithm and model.A MP2P super-node selection algorithm based on P-AHP (Personalized Analytic Hierarchy Process) is proposed. The storage capacity, computing power, the robustness, the energy conservation and the security are selected as the five KPI (Key Performance Indicators). The node comprehensive performance evaluation can be obtained by using the AHP (Analytic Hierarchy Process), and realizes the MP2P super-node selection. The weight of KPI can be personalized adjusted according to the needs of users business, which embodys the flexibility of the P-AHP algorithm. Simulation results verify the effectiveness of the proposed algorithm.A MP2P network QoS adaptive guarantee model (NGDA) based on neural network and goal-driven is proposed. Services are divided into two types of preferred services and alternative ones, the BP neural network detects the current service QoS in real-time; when the service QoS does not meet the needs of users, NGDA model can adaptively chose an alternative service meeting requirement to replace the original one. The personalized service customization reflects dynamic and flexible features of the NGDA; the whole model runs in a heterogeneous network environment. Simulation based on OMNeT++ show that NGDA model has a certain improvement in service request success rate, which overcomes the impact on services QoS caused by the inherent deficiencies of MP2P network.In this paper, super node selection, resource discovery, traffic management, trust model, QoS security and other MP2P key technologies are deeply researched in heterogeneous network environment. Then, a number of novel algorithms, models and management mechanisms are proposed. In addition, by dint of mathematical modeling, theoretical analysis and OMNeT++ simulation analysis, the proposed algorithm and model is verified to be effective; meanwhile, the future research directions are also predicted.