Video Transcoding And Video Recommending Based Caching Design

Cache algorithm is a key factor that affects the cache performance of cache server(s).Traditional caching algorithms usually measure the value of a video based on the popularity of the video file and preferentially buffer the most valuable videos.In recent years,due to the diversification of terminal devices,a video usually needs to provide different bitrate versions to adapt the demands of different terminal users.Since such video versions can be converted by video encoding technology,it provides opportunities for further optimization of caching design.For example,Scalable Video Coding,SVC has been used for caching design to improve caching performance by caching the most popular video layers.On the other hand,as the video contents are greatly enriched and recommendation technologies,more and more users are accustoming to accept the recommended videos to watch,such that video recommendation can also be utilized to design caching to improve caching performance.Different from the existing researches,on the one hand,considering the better compatibility of video transcoding technologies,we proposes a video transcoding based caching model,which enables the cache server to buffer at most one valuable version for each individual video to improve caching performance;on the other hand,considering that the existing video encoding based and video recommending based caching are independently studied,we propose to jointly use the two types of technologies to design cache,to further improve the performance.Concretely,the contributions are as follows.(1)This paper establishes a transcoding based cache optimization model and proposes a corresponding cache algorithm to reduce video delivery delay.The algorithm defines the value of a to-be-cached file based on the further reduction in delivery delay considering that once it is selected to buffer it can satisfy more users' viewing requests and thus delivery delay can be further reduced.The algorithm then iteratively finds the most valuable video versions to cache in a greedy manner,one version each iteration,until no video files can be put into the cache system.Simulation results show that the proposed transcoding based caching algorithm can reduce the average delivery delay by up to 40%,much higher than the traditional algorithm.(2)This paper establishes a joint caching optimization model based on both video transcoding and recommending and proposes a corresponding heuristic algorithm to improve caching performance.We define the value of a video version based on the further delivery delay reduction benefited due to the further more satisfied users' demands by this video version if it is buffered.We then propose the algorithm to iteratively find the most valuable video version to buffer in a greedy manner.Simulation results show that the proposed algorithm has significant better caching performance when the cache size is relative small.