Video Caching Schemes In Edge Networks

With the development of mobile communication, social networks and mobile intelligent terminals, people have more demand for high-quality real-time data services, especially online high-definition videos and social contents. The traditional Content Delivery Network (CDN) is far from users in edge networks and this result in a long transmission distance and delay. However, the trend of business localization is more obvious, especially social networks make content delivery more marginalized and users combined to small circle in edge networks.Therefore, caching in edge networks has attracted widely attention. In this thesis,we classify video content delivery services into traditional video services where users only consume video services and social video services where users may actively produce or forward videos. In this paper, we focus users’ video content preferences and social relationships to design caching schemes in edge networks.This thesis includes the following two aspects.Firstly,designing a caching scheme in edge networks based on users’ video content preferences. If users’ video content preferences are used to design caching schemes, the number of video contents is an important factor affecting the complexity of caching schemes. Caching schemes previous research considered is about content-level and this makes caching schemes rather complex and cannot be practical. In order to reduce complexity, we first apply clustering algorithm to cluster users into groups and then divide the video content space into subspaces in order to reduce user dimensions and video content dimensions simultaneously.Then, a greedy caching scheme is proposed which reduces the complexity from exponential or high-order polynomials to quadratic with the number of video contents and the number of users. Moreover, this paper presents some new characteristics of video requests caused by personalized video recommendation systems and generates synthetic data to imitate them. Then we evaluate their impacts on actual performances of caching schemes using different clustering algorithms, especially when the file preferences matrix about users video file requests becomes high dimensional, sparse and high-asymmetric.Secondly, designing a caching scheme in edge networks based on users’ social relationships. In this paper, we add social relationships as an important factor in modeling caching problems. In order to deal with mobility of users in edge networks, we combine virtual social relationships and physical social relationships. In additional, we model users’ movement based on their social relationships. Afterward, we propose a two-step caching scheme, specifically, we firstly choose the caching nodes according to the strength of social relationships and then decide where to cache and what to cache in the chose caching nodes. The simulation results show that the video content caching scheme based on social relationships is superior to the traditional video content caching scheme based on video content popularity in social networks. In addition, the effect of physical and virtual social relationships on designing video content caching schemes is studied in the thesis and due to the randomness and dynamics of user movements; the physical social relationship has stronger influence than the virtual social relationships on video content caching scheme.Finally, a brief summary of this thesis is made and possible future research contents are point out.