Research On File-Sharing Service Optimization Technology Of P2P Network

As a novel Internet application, Peer-to-Peer Network (P2P) provides a file sharing solution with features like scalability, high security and easy deployment by changing network organizational mechanism, thus fulfilling the progressive demands of the Internet. In the past few decades, researchers have analyzed file sharing service in both aspects, say P2P network topology and the controllability of the transmission, and have gained quite an amount of achievements in various fields such as structural topology, redundant distribution strategy, fragmentation mechanism, etc. However, the researches mentioned above simply considers the nodes in the file sharing system as computers with ability to compute and communicate, and ignores the impact of real world user behavior upon service, therefore the problems such as free-riding, jitter and inefficient file search would arise. Therefore, it is an important research subject to explore the relationship between user behavior and file sharing P2P network services, and then seek for optimization upon corresponding relations.This paper starts with the user behavior analysis of the P2P file sharing network, finds the influence of properties such as aggregation and diversified service upon application services, then explores the optimization technologies with the aid of these features, and makes an intensive study of the network structure suitable for the optimization technologies. The research findings are as follows:1.An analysis on file sharing P2P network is done by the presentedmulti-angel analysis model in the following aspects:resource, user and network. With the combination of the conception that file sharing P2P network is the application layer virtual network consists of users and relative resource objects, such an analysis will be able to reflect the basic situations of application service and user behavior clearly, and a solid statistical foundation can thus be provided for file sharing optimization model and relative technologies.2. A model called AF-IChord is carried out taking advantage of the correlation of quantified user interest versus user topology in order to solve the inadequacy in the usage of semantic information from users with high network clustering coefficient. The AF-IChord model classifies user with the similarity of interest and online law, and renews a redundant item in the routing table into the user category information which combines the interest and network usage pattern, thus improves the search and access efficiency of file sharing P2P network.3. A node service availability-based resource scheduling algorithm called APPF is carried out in order to solve the stability problem of the file sharing service as a result of ability discrepancy of various users. APPF uses the network availability and user availability to indicate the node resource service availability. Network availability takes DNS and IP as static measurement in order to reduce the topology blindness in service node selection, and user availability uses the service attribute of the node to effectively discover the user role information. APPF labels the high service availability nodes in the file sharing P2P networks based on those meta data, thereby succeeds in the schedule of advantageous resources. With comparison to random walk scheduling algorithm, APPF can effectively lessen the influence of free-riding users and users with unstable connections upon the network services.4. Based on the consideration that the optimization technologies based on user classification and node service availability such as AF-ICHORD and APPF are still not suitable for existing popular P2P network topologies, a heterogeneous protocol hierarchical model called HCC is carried out. HCC uses CAN protocol in the super node layer and Chord protocol in the normal node layer, so there is no need to maintain the coherency of the attributes in both layers, then the fault tolerance capacity becomes better while the maintenance cost is reduced. Besides that, the identifier for super node management can be represented into multiple dimensions, so the load of each dimension can be calculated and managed separately, and a fairness index is introduced to adjust the payloads among super nodes effectively.