Research On Cache Management Of HTTP Adaptive Streaming Media On Demand Cloud Platform

With the explosive growth of intelligent terminal devices in the Internet,a massive of streaming media video on demand services not only promote the development of HTTP adaptive streaming media technology,but also bring great challenges to cloud platform services.The existing intelligent terminals can be divided into fixed terminals(such as smart TV,host computer,etc.)and mobile terminals(such as smart phone,portable computer,etc.).Compared with mobile terminals,fixed terminals have larger storage capacity,more sufficient power and more stable bandwidth.This kind of terminal equipment can provide shared services while providing its own services,but it is not easy to carry.The uncertainty of the physical access point of the mobile terminal may cause the streaming video on demand service to need to switch the server many times,which will greatly affect the user viewing experience.The cache management is regarded as one of the key technologies to ensure the user experience.The best experience of users is that the local server caches the requested target video in advance.However,often the same cache strategy could not well guarantee the experience of various terminal users.Therefore,this paper studies the cache management of HTTP adaptive streaming media on demand cloud platform for the services of fixed terminals and mobile terminals respectively.The main work is as follows:Firstly,this dissertation constructed a CDN-P2 P hybrid network service architecture,and proposed a cooperative caching strategy based on optimal utility.The architecture consists of two layers of overlay networks.One layer is composed of CDN peers and the other layer is composed of super peer groups.CDN peers are deployed by ISPs,and super peers come from a massive of fixed terminals.First,in order to select super peers,a peer performance evaluation algorithm is proposed.Considering the service capability limitation of a super peer,a super peer group self-organization algorithm is also designed.Then,this dissertation analyzes the supply-demand relationship between ISP domains and constructs a utility model.And,the utility model is optimized through the greedy algorithm.Finally,a cooperative caching strategy is proposed based on the utility optimal model.The simulation performance of the proposed cooperative caching strategy is verified by the experimental results that the cooperative caching strategy can reduce the cost of cross ISP domain traffic transmission and improve users’ QoE(Quality of Experience).Second,this dissertation constructed an edge network service architecture in a mobile user environment,and proposed a cache replacement algorithm based on local video popularity and video size.First,in order to effectively solve the dynamic problem of video heat,this dissertation constructs a local video heat model,which is divided into heat attenuation model and heat rise model.In addition,considering the problem of a sudden reduction of video popularity caused by very few access frequencies in each time period,the attenuation model can be further divided into frequency dependent attenuation model and frequency independent attenuation model.Then,information entropy is introduced to analyze the weight of local video heat and video size in real time.Finally,experiments show that the proposed cache replacement strategy can effectively improve the cache hit rate and reduce the backhaul link load.