节点文献
移动P2P网络中基于方向搜索算法的研究
Research on Direction-Based Search Algorithm in Mobile P2P Network
【作者】 刘孝男;
【导师】 房至一;
【作者基本信息】 吉林大学 , 计算机系统结构, 2010, 博士
【摘要】 移动P2P是一种新型的资源共享和数据分发的手段与技术,移动应用环境的逐渐成熟,使其拥有越来越广阔的应用前景。移动P2P网络是由互相通信的移动设备组成的,与有限网络上的P2P系统比较而言,移动P2P有着更多的约束。由于电池能源、无线宽带的限制以及动态变化的网络拓扑结构,移动P2P对路由、资源发现、数据检索、安全和隐私管理等提出了新的挑战。洪泛策略是分布式非结构化P2P网络中最流行的资源搜索机制,但该机制在处理每个节点的搜索消息时需花费昂贵的电池能源,带宽和计算资源;分布式哈希表DHT是用于结构化P2P网络中配置和搜索数据资源的,但该策略的使用需要事先建立搜索索引,限制了灵活性。上述的方法都无法很好的执行与移动P2P网络。因此,研究者们为了能使P2P技术在移动无线网络中得到充分的利用做了大量的研发工作。如果能将基于洪泛式搜索算法的改进与基于结构化分布式DHT搜索算法的改进有机地结合起来,通过超级节点和普通节点将基于DHT的搜索机制与洪泛机制相结合,使用较少的副本、缓存或控制信息,尽量减少洪泛的次数,充分利用上述两者的优点,将是资源搜索在今后的一个发展方向。本文作者提出了基于方向的搜索算法-DBS算法,该算法秉承了Flooding算法及随机行走算法的优点,同时又解决了Flooding算法及随机行走算法自身存在的缺点,通过模拟实验得出的实验结果验证了DBS算法在搜索过程中的有效性及可操作性;设计了一种移动P2P覆盖网络,为后续设计的算法提供了一个合理的运行环境;基于DBS算法思想设计了适用于移动P2P网络的MDBS算法,MDBS算法包括基于代理的方向搜索算法-ADBS算法和动态调整转发方向搜索算法-DTDBS算法。选择MATLAB为本仿真工具,模拟了MDBS算法。通过仿真实验,首先验证了算法运行所在覆盖网络的有效性,其次通过度量标准对分别对ADBS算法和DTDBS算法的模拟结果进行了详尽的分析,最后通过性能分析更进一步验证了MDBS算法在移动P2P网络中的有效性及可行性。
【Abstract】 P2P technology which is a high effective and directly information transmission mode, it breaks the traditional C/S pattern, makes the messages exchanging between peer-to-peer equipments and gets widely used in the internet. With the development of transferring data service and the popularization of load bearing data service terminal, mobile terminal and network integrate more and more closely, how to realize the peer-to-peer information shared has taken attention widely.Mobile P2P network is comprised of many mini-type equipments. With the category and quantitative of mobile equipments increasing continuously, more and more users also continuously pursue new applications taken by mobile equipments. Due to the special nature of mobile devices lies in its limited bandwidth, low storage and processing power, connection stability and short battery life-cycle, as well as unpredictable connectivity and topological dynamics, etc. Therefore, effectively search and locate resources are very important in mobile P2P network. Traditional P2P network can be divided into centralized P2P network, distributed P2P network and hybrid P2P network, distributed P2P network is divided into distributed unstructured P2P network and distributed structured P2P network.Centralized P2P network, there is a central directory server, which is for all nodes of P2P network to provide services of search and share files; distributed unstructured P2P network uses completely random flooding-style search and random transmitting mechanism, representative search technologies such as Flooding search algorithm, Modified-BFS search algorithm, Random Walk search algorithm, Gnutella2 search algorithm, and Mobile Agent-Based search algorithm, etc.; distributed hash table DHT is the resource position method taken by the most structured P2P networks, the classic DHT search algorithm includes Chord, CAN, Pastry, and Tapestry; application cases of typical search algorithm in Hybrid P2P network include Kazaa, eDonkey, and eMule.P2P business models which are based on mobile terminals are mainly two ways: First, mobile terminals connect to a traditional P2P network system through the mobile network to provide P2P technical supports for mobile users. In this business model, you need to use P2P network architectures and algorithms of traditional networks in the key technologies such as network architecture, resource search location according to the characteristics of mobile terminals. Second, the mobile terminals are in close proximity may form Ad-Hoc Networks, achieve P2P overlay networks in Ad-Hoc networks, namely mobile P2P networks for sharing resources. According to the coverage networks whether have agents or not, mobile P2P system is divided into two categories: agent-based mobile P2P systems and mobile P2P systems without agents. For the limitation to mobile terminal resources, most of nodes are connected to networks with agents. A few systems adopt lightweight strategies that would enable mobile terminals directly access to the network without agents.According to topology structures of overlay networks, mobile P2P network structures can be divided into centralized network architecture, semi-distributed network architecture, and full-distributed network architecture based on mobile Ad-Hoc network. In the centralized network architecture, researchers put forward some search algorithms such as the eDonkey protocol is used in mobile networks, and the idea is realized in GPRS networks and UMTS networks. In the semi-distributed network architecture, researchers put forward some search algorithms such as the search algorithm which is based on flooding-type messages supports the information sharing services of mobile terminals through raising mobile agents in the sharing system between Gnutella networks, and the search algorithm which is based on semi-distributed DHT takes the super-node as the basis of the improved algorithm to CAN. In the full-distributed network architecture based on mobile Ad-Hoc network, researchers put forward some improved search algorithms such as 7DS, RAON, Konark, MPP, and ORION which are based on flooding-type search, At the same time, some researchers put forward some improved search algorithms which are based on the structured distributed hash table, For example, routing strategy based on location, sharing P2P video streaming by using AP as a resource search in mobile Ad-Hoc network environment, using vehicles to be super nodes in a mobile P2P network, while serve as control nodes, composite the backbone network of mobile P2P systems and so on.Search algorithms applied in mobile P2P networks and their corresponding improved algorithms although they are each with its own features, but there are still some limited defects such as increasing of network redundancy traffics, increasing of network load, and flexibility constrained. Therefore, researchers do a lot of researches and developments in order to enable P2P technology can be fully utilized in mobile wireless networks. How to make the network redundancy traffic decrease, reduce the network load, give full play to the search flexibility will be a research focus on resource search of mobile P2P network in the future. Based on the above thought, we research the search algorithm applied in mobile P2P network in this paper, specifically, the main contributions are summarized below:First, propose the direction-based search algorithm-DBS algorithm according to the design conception of Flooding algorithm. Describe the basic idea and the detailed design process of DBS algorithm, and take simulation experiments for DBS algorithm, carry on data analysis to the experiment results obtained from simulation experiments. First of all, by setting a definition, a lemma and a theorem prove DBS algorithm has the connectivity in the operation space. Second, by the analysis to the two experimental results of digital datum and image datum fully reflect effectiveness and feasibility of the algorithm in the search process. DBS algorithm implements the search process through the program to simulate P2P networks in a two-dimensional space, done with the advantages of Flooding algorithm and its improved algorithm, such as low requirements and simple implementations to the network topology, support complex resource search and effectively avoid sending repetitive messages and reduce generating redundancy messages, while ensuring coverage of the whole search, all that establishes the foundation for the improvement to follow-up search algorithms. Secondly, design one mobile P2P overlay network, the overlay network adopts the overlay abstract design method of integrated and structured for constructing abstract overlay to P2P system in mobile Ad-Hoc network, combines the relationship of mobile P2P network and mobile Ad-Hoc network to draw abstract overlay map of P2P network based on mobile Ad-Hoc network. In the overlay network we set up logical space domain and physical space domain, the mobile Ad-Hoc network layer which logical domain corresponding to is the mobile layer, the fixed P2P network layer which physical domain corresponding to is the fixed layer, nodes on the fixed layer has the determined position, thus the fixed nodes form a network information database on the fixed layer, the database is used to find and store the state of the node with the same interest. Nodes on the mobile layer can always change their position information in the network level, while mobile nodes can always send and receive position updates from each other. The fixed layer adopts structured P2P system based on distributed hash table, its main task is to deal with the mobility of mobile nodes and the resolution of the address, in order to achieve exchanging between mobile P2P network and fixed P2P network, the overlay network provides a reasonable operating environment for MDBS algorithm.Thirdly, design MDBS algorithm applied for mobile P2P network, MDBS algorithm includes agent-based direction search algorithm-ADBS algorithm and dynamic transmission direction search algorithm-DTDBS algorithm. The algorithm combines the improved algorithms which are separately based on flooding search algorithm and structured distributed hash table search algorithm, combines the search mechanism based on DHT and flooding mechanism through super nodes and ordinary nodes. In ADBS algorithm, a super-node is set on the mobile layer, one agent node is set in each Ad-Hoc network, the super-node is responsible for the coordinates of each Ad-Hoc network, storing the information of node packets and the management to agent nodes on the mobile layer, the agent node is responsible for storing messages from neighbor Ad-Hoc networks and transferring messages within and between Ad-Hoc networks. If the position of the agent node on the mobile layer is under the case of unknown or invalid, the nodes within the P2P network on the fixed layer will resolve the address for the agent node, allows the search to proceed smoothly. DTDBS algorithm sets the connection network and the direct transiting network, compared with ADBS algorithm improves the search speed of the network overlay. DTDBS algorithm sets Dynamic Table, Collection Table, life-cycle of search information and interest-related degrees introduced, compared with ADBS algorithm shortens the response time, controls the time of messages transited, reduce the generation of redundant messages, DTDBS algorithm further optimizes ADBS algorithm.Fourth, chose MATLAB as the simulation tool, through setting up a simulation environment and setting the simulation parameters to simulate MDBS algorithm. Through simulation experiments, we test scalability and routing efficiency of the overlay network, the results show the load of fixed nodes do not increase with the increasing of the mobile node quantity when the algorithm runs, so the scalability to the overlay network is ensured; as the mobile nodes in the proportion of the total nodes gradually increase, the average hops of routing do not change a lot, thus ensuring the efficiency of routing. We carry on simulation tests to ADBS algorithm and DTDBS algorithm separately through measure standards, the test results of ADBS algorithm show that the algorithm has the scalability because the traffic quantity of average messages increase with the increasing of node quantity; the average search path lengths correspond to the actual network operation with the changes of mobile node quantity, the search traffics of nodes are adapt to the changes in the search speed of mobile nodes. The test results of DTDBS algorithm show that the algorithm sets the life-cycle of search messages and the direct transiting network, introduces the interest-related degrees, compared with ADBS algorithm further shortens the search response time, gives the search feedback of users in a shorter period of time, improves the customer satisfaction effectively. Finally,through performance analysis we further verify the validity and feasibility of MDBS algorithm in mobile P2P network.