Research And Implementation Of Content Forecasting Mechanism In CDN-P2P System Based On Hadoop

In recent more than ten years, with the rapid development of the internet, network information and number of users have increased dramatically. The contents which are shared and transferred in the network have been extended from simple text and picture to audio, video and some complicated and various multimedia. In order to efficiently delivery content, alleviate network congestion and improve user experience, CDN and P2P technology, as the main network content distribution technology, have been widely used in many fields. Considering the complementation differentiation between P2P and CDN when they provide content service, the technology integrating P2P with CDN is becoming a hotspot research field.As the network scale is growing larger continuously and share resource information is booming, more problems related to the file sharing among the peers and to the file providing by surrogate are coming. Main show is the requirement to load capacity of surrogate server storage and the limitation of time expending for the response of file request by the peer. With the increasingly growth of nodes that need sharing service and number of files provided, files should be frequently transferred between surrogates or between surrogate and central server. So, not only time expending for the response of file request among peers increases, but also bandwidth resources are consumed. At the same time, users also need to spend a lot of time to find the content that users need in mass resource information.Improving the content cache strategy of surrogates in the CDN-P2P network, quickly responding to file request of node, raising efficiency of finding requirement content and sharing content for node users in the mass information, are the important research development direction of CDN-P2P technology in the future. By analyzing features of CDN-P2P network, especially influence factor of initiative participation for node user, intelligence recommendation and search engine technology are applied to CDN-P2P prototype system based on the Hadoop in this paper.The research points of this paper are listed as follows:(1) By analyzing the relation between type attribute of sharing content and node requirement, the factor of user preference is calculated. According to rating similarity among peers and user preference factor, prediction function in the collaborative filtering method is improved, and the paper researches and analyzes node requirement predication model.(2) Traditional CDN technology is studied. Combined with subnet organization characteristics and similarity among peers in current CDN-P2P system, content storing strategy in the system is designed again.(3) In view of content sharing requirement and information searching needs for node users, a search system which is used to help share content is designed and implemented by adopting Solr. In this search system, users could retrieve resource information through inputting key words.(4) Node requirement predication model and content storing strategy of surrogate server that are proposed above are implemented in the CDN-P2P prototype system.