A Study On Cooperational Caching Algorithm In CDN

Online video viewing is developing rapidly. Global video traffic has accounted for more than 70% of the Internet traffic. What’s more, with the development of the popularity of smart phones and wireless technology, the ratio mobile video traffic accounted for the Internet traffic is close to two thirds. Traffic load is very large. Caching the popular video content based on the popular video of the long tail distribution characteristics is commonly used for the ISP to save bandwidth. Optimizing the cache strategy based on the analysis of the user and video characteristic is the focus of this research. The main work of this paper is as follows.Firstly, we analyzed the Sina video data from two aspects viewing behavior and character of video popular. From the aspect of user viewing behavior, There are two viewing peak in the afternoon and evening. From the aspect of popular degree, It can be summarized into three points:first,20% of the video occupied the 80% traffic; second, popularity of video is the same in different regions and different ISP; third, the video popular time can be maintained for one day.Secondly, we established two models based on the analysis of the user viewing behavior and video popular degree. The result of cooperation cache can save 83.28% traffic than the independent cache. But the model complexity is very high. In this paper, we introduce the clustering method to reducing the capacity. And the solution time reduced to 0.183s. But during the process of cluster, the redundant is increase. So we introduces the algorithm of removing the redundant. Finally, the time remain the same level while the traffic is saved by 81.36%.Thirdly, since the popularity of video is changing by time, the traffic saved decrease. The general update algorithm mainly includes two kinds of offline update and online update. The offline update algorithm is close to the optimize solution but with high complexity. Online update algorithm complexity is small but with worse performance. So in this paper, we put forward an online update algorithm cooperating with an offline update algorithm. The ability to save bandwidth is stable and effective throughout the update cycle.Fourthly, in the large users of mobile access environment (e.g. Stadium), management ability and bandwidth capacity of the base station is not enough. Since the D2D transmission ability is strong, we presents the D2D auxiliary content distribution algorithm. Through NS3 simulation, the average time of all user meeting their needs reduce to 29.06%.