Research On Content Prefetching For Distributed Content Dilivery Network

Obtaining content from the edge of the network is an important idea for the Content Delivery Network(CDN)to speed up users' access to high-bandwidth content such as large files,high-definition video,which can reduce the access pressure of the central server.However,the expensive deployment cost and management cost make the expansion of CDN very difficult.CDN-P2 P technology reduces the dependence of content on CDN by referring to Peer to Peer technology(P2P),which is an important way for CDN distributed deployment.One of the important features of CDN-P2 P is to increase the efficiency of content distribution by caching content on large-scale distributed user nodes.However,there is a certain lag in caching technology,that is,content caching needs to be triggered by user requests.With the employment of prefetching technology,the requested content of users can be pre-cached,and the response latency of user requests can be reduced.It is often used as a supplement to caching technology.However,the existing CDN-P2 P prefetching technology neglects the relevance between content resources and topology,resulting in unreasonable location distribution of prefetched content,which not only increases the redundancy of prefetched content in the network,but also causes the problem of low efficiency of content sharing among nodes in the topology.In order to solve the above problems,the thesis firstly proposes a community network based on interest network model by using user interest model and information interaction to promote content sharing between nodes and improve the cache hit rate.Secondly,on the premise of realizing the interest network model,a reasonable prefetching strategy is formulated to reduce the redundancy of prefetched content,and improve the accuracy of prefetched content.(1)A capability model is constructed for the nodes according to the physical capabilities and stability of the user nodes and the mathematical model-preference vector for the interests of users and communities is determined based on the interest factors of the nodes on different content topics.In addition,by improving the cohesive hierarchical clustering algorithm,a method for constructing CDN-P2 P network topology is obtained.The method combines two communities with similar interests by exchanging information between user nodes,and finally obtains a network topology based on interest communities.(2)By analyzing the Ebbinghaus forgetting curve,the interest attenuation factor is obtained,and the user's preference for the topic is updated incrementally to get the user's real-time interest model.According to the interest model,a group recommendation method suitable for the interest community is proposed,and the prefetched content recommendation list is generated according to the recommendation method.Through defining the the value of prefetched content and cached content,content prefetching will be performed according to the prefetched value.(3)The OMNET++ simulator is used to validate the network simulation.The experimental results show that the topology construction method based on interest community can complete the topology construction in a relatively short time.Compared with the prefetching strategy based on popularity,the prefetching strategy in interest community can improve the accuracy of prefetching content by about 15% and the hit ratio of the system by about 15%.