【作者】 李军；

【导师】 张国印；

【作者基本信息】 哈尔滨工程大学 ， 计算机应用技术， 2013， 博士

【摘要】 随着移动网络的迅速发展和移动设备的快速普及，对等网络技术在移动网络中的需求越来越强烈。移动对等覆盖网能够屏蔽网络路由协议和物理网络设备标识、编址之间的差异，提供相对稳定的拓扑结构和路由、负载均衡、资源查询等服务。移动对等覆盖网的结构直接决定了移动P2P系统的可扩展性、鲁棒性、安全性和抗扰动性。因此，对于移动对等覆盖网的研究具有十分重要的意义和应用价值。虽然已有一些移动对等覆盖网被提出，但其性能和抗扰动性较差，难以得到实际应用。本文在现有研究的基础上，针对移动对等覆盖网研究中存在的一些问题，对移动对等覆盖网的构建、评估、优化等关键技术进行了深入的研究。本文通过分析多种经典对等覆盖网的基本原理和拓扑结构，深入剖析了传统对等覆盖网不能完全适应移动网络的原因，对移动对等覆盖网的概念进行了明确的阐述，并对现有的移动对等覆盖网进行了详细的比较和分析，为实现移动对等覆盖网的方法验证和性能评价，对多种网络模拟器进行了比较和选择。在上述研究基础上，本文首先提出一种抗扰动的移动对等覆盖网，根据节点加入网络的先后顺序构造多个不同k值的k-派系社区结构，并通过外联节点来链接不同的k-派系，针对节点的加入及k-派系的形成、节点退出、节点失效等情况设计相应的算法。同时，通过网络拓扑控制过程和外联过程来不断调整覆盖网拓扑结构,实现对移动物理网络拓扑变化的自适应，并结合数据冗余和失效节点主动检测来提高移动对等覆盖网的抗扰动能力。其次，为实现拓扑控制信息的有效传播和保持节点资源列表的副本一致性，本文对移动对等覆盖网中的数据分发机制进行了研究，根据移动网络节点资源受限和网络链接不稳定的特点选择合适的数据分发方法，并基于移动对等覆盖网的拓扑结构进行改进。通过采取对不同状态节点动态调节分发概率的策略，增加不稳定节点的分发概率，保证数据能够在较短时间内分发到所有目标节点，提高系统在扰动情况下的数据分发效率，有效降低网络负载。再次，本文对移动对等覆盖网在扰动情况下的性能评价体系进行构建，并遵循这一评价体系，在指数扰动模型和KAD扰动模型下对多种覆盖网进行性能评估。本文进一步对基于频度度量的扰动评估模型进行研究，通过加入影响网络扰动的间接因素使扰动评估模型能够更加真实的反映实际的网络扰动情况。结合模拟器的实验结果和多项式曲线拟合方法，本文提出一个适合任何移动对等覆盖网使用的估算扰动评估公式参数权值的方法。最后，本文基于改进的遗传算法对移动对等覆盖网络参数进行自动模拟优化，以提高移动对等覆盖网络的性能。为了克服经典遗传算法的全局优化效率差和易发生“早熟”现象的缺点，通过计算个体之间的相似度来减少不必要的交叉操作，并提高相似度低的个体之间进行交叉操作的概率。利用自动模拟优化算法来实现对模拟器的配置文件修改、源代码修改和编译、运行等过程的全自动运行，从而提高优化的效率和可靠性。本文的研究工作围绕移动对等覆盖网的关键技术展开，其中涉及到的网络构建、数据分发、性能评估以及性能优化为今后的研究工作奠定了良好的理论基础和实验数据基础。更多还原

【Abstract】 The demand for mobile peer-to-peer technology becomes much stronger because mobilenetworks are developed quickly, and mobile equipment and applications are more widely used.Mobile P2P overlays can shield the difference among the network routing protocols and themarks and addresses of physical equipment, and provide services for stable topology, loadbalance, and resources search etc. The structure of mobile P2P overlays directly determinesthe scalability, robust, security, and churn-resistance of the mobile P2P systems. Therefore theresearch of mobile P2P overlays is of great significance and application value. Although therehave been some mobile peer-to-peer overlays proposed, they get poor performance andchurn-resistance, which make them difficult to get practical application. Based on the existingresearch results, construction, evaluation, and optimization are intensively studied to solvesome problems in the research of mobile P2P overlays.By analyzing a variety of traditional peer-to-peer overlays with the basic principle andtopology structure, this thesis further expounds why the traditional P2P overlays can notcompletely adapt to the mobile networks. The concept of mobile P2P overlays is illustrated,and the existing mobile peer-to-peer overlays are analyzed and compared in detail. Thenetwork simulators are compared and chosen for the method validation and performanceevaluation of mobile P2P overlays.Firstly, the thesis proposes a mobile P2P overlay under the basis of the above mentionedresearch. The overlay constructs k-clique communities with different k value according to theorder of each node joining the overlay, uses contact nodes to link distinct k-cliquecommunities, and designs corresponding algorithms for joining, leaving and crashing of thenodes and construction of k-cliques. Meanwhile, it keeps adjusting the topology of the overlayby using topology control process and inter-clique connection processes to improveself-adaption for change of physical network topology and to increase the churn-resistanceability by combining data redundancy and proactive detection of failure nodes.Secondly, in order to realize the effective transmission of topology control informationand to maintain replica consistency of the node resource lists, the thesis researches on the datadissemination mechanism based on mobile P2P overlays. The appropriate data dissemination method is chosen according to the characteristics as resource-constrained nodes and linkinstability in mobile networks, and is improved based on the mobile P2P overlay topologystructure. By taking on the strategy as dynamic adjusting the dissemination probability ofnodes with different state, the dissemination probability of unstable nodes are added, whichmakes sure that the data can be disseminated to all the target nodes in a relatively short periodof time, the efficiency of data dissemination under churn is enhanced, and the network load iseffectively reduced.Thirdly, the performance evaluation model of mobile P2P overlays under churn isconstructed. Follow the model, performance of several overlays are evaluated under bothexponential churn model and KAD churn model. Furthermore, how to improve the existingchurn evaluation model based on frequency measurement is studied; by adding the indirectfactors influencing network churn, the improved churn evaluation model can more reallyreflect the actual network churn situation. By combination of simulation results andpolynomial fitting method, this thesis studies the parameter weights estimation method whichcan be used in any mobile P2P overlays.Lastly, parameters of mobile P2P overlays are optimized based on improved geneticalgorithm and automatic simulation to improve the performance of mobile peer-to-peeroverlays. In order to overcome the poor efficiency of the global optimization and classical"premature" phenomenon occurred in genetic algorithm, unnecessary crossover operations arereduced by computing the degree of similarity between individuals, and individuals with lowersimilarity degree get the greater probability of crossover operation between them. Automaticsimulation optimization is used to modify the simulator configuration file and the source codeand to compile and run automatically so as to improve the efficiency and reliability of theoptimization.Research of this thesis addresses the key technologies of mobile peer-to-peer network,which involves the network construction, data dissemination, performance evaluation andoptimization of network parameters, these researches lays a good theoretical basis andexperimental data base for future studies.更多还原