QoE-based Distribution Policy For Content Distribution Networks

Nowadays,CDN server primarily aims to guarantee QoS,such as latency,bandwidth,loss rate,etc.CDN can alleviate server pressure and improve the website performance and scalability.With requirement of stream media,it is difficult to meet user demand for services subjective perception relying solely on the original CDN distribution policy.The users' subjective perception is depend on user experience for something.Once the new thing can bring larger user experience than original thing,the user will give up the original thing and turn to the new thing.Omitted from the end-user point of view to improve the quality of service of CDN,resulting in a phenomenon of a large number of user are lost.So it is crucial to Improved distribution policy to enhance the user for service-aware.From the view of QoE,this article study about the CDN distribution strategy,to improve the user experience of the network.The main idea of this algorithm is based on the user interest to guide the CDN content distribution.Therefore we need to extract the users' behavior from the web log file and abstract user intersect based on the behavior of users.Firstly,the web log file should be preprocessed to get valid user data.Identification the user session sequence from the preprocessed log file and calculating most frequently accessed pages by user are needed to do.Meanwhile the markov chains model and k-means algorithm is used to cluster session into groups and get the user interest model.According to user public interest and page frequent item sets,CDN distribute content to surrogate server.Moreover,this article also take into accout the limit capability of surrogate server.The method of measure QoE include statistics and psychology subjective evaluation,objective evaluation and a combination of subjective and objective method.Due to the subjective evaluation methods have too many restrictions in practice,so this article choose objective evaluation method and use four objective index(the number of hits,response time,byte hit ratio and throughput)to evaluate it.Finally,the CDNSim is used to simulate the CDN environment and compare with original strategies to verify the algorithm of this paper.