Research On Video Cache Updating And Prediction In Mobile Edge Computing

With the popularity of various video social platforms,such as Tik Tok,Bilibili and Kuaishou,huge data flows will be generated in the network transmission links.Under the traditional cloud computing model,the user's request is to deliver content in the cloud,and cloud computing does not have the characteristics of low delay and high speed in the case of huge data flow.Mobile edge computing is to sink the computing and storage resources to the edge close to the user,and the edge will handle the user's request,which will make up for the shortage of cloud computing.The popular video data content can be cached in MEC,and the content delivery by MEC will reduce the content delivery delay,improve the network bandwidth capacity,and reduce the network link pressure.In addition,in the edge cache,with the development of the recommendation algorithm,the user's access behavior becomes predictable,so the video data that users may access can also be cached to the MEC node for storage,so that more users' video requests can be processed at the edge.Some progress has been made in caching and prediction in MEC,but it is not accurate enough in measuring video popularity,and it fails to combine active caching with reactive caching.In view of the above analysis,the main work is as follows:First,by analyzing the characteristics of video,this thesis proposes a reactive cache update algorithm to measure the popularity of video,which comprehensively considers the change of video with time,with the change of access frequency and with the change of video quality.Then,through the crawler technology,the video access information of users on Bilibili is retrieved,such as the amount of play,date of video release,thumb up number,etc.Based on the crawling data,the proposed algorithm is used to calculate the popularity of the video.When the MEC node is full,the video with high popularity is replaced with the video with low popularity.The results show that the algorithm can improve the cache hit rate.At the same time,different attenuation coefficient will also have an impact on the hit rate.Secondly,based on the reactive cache algorithm,a cache prediction system based on the MEC cache architecture is proposed.The improved collaborative filtering algorithm is used to advance the video that users may watch and the video is not stored at the edge of the cloud cache to the edge storage.Through the experiment,the impact of different prediction coefficients and video play ratio on the hit rate is analyzed.The final results show that the algorithm can further improve the cache hit rate.