Caching Based On Actual User Behavior

In the Cable Digital Television System, it needs high speed of reading and writing. The bandwith of the disc I/O and the bandwith of the network may become the bottleneck of the system. The main issues include server overload, high possibility of network congestion, deteriorated user experiences, for instance due to delay of application start time, and etc. Caching is an effective method to solve these problems.Performance of caching algorithm is highly related to the way and the content of user interactions.This article proposed a high- capacity caching algorithm based on the analysis of user behaviors. The contributions include separation of the cache into three parts, for storing prefix, suffix and common component respectively, as well as design of access control and replacement strategy. This article approach utilizes linear subsection method and a value function that emphasizes instant messages while also taking into account historical information. The real-data-based simulation experiments demonstrate the algorithm achieves higher byte hit ratio, compare with exponential segmentation algorithm and the adaptive and lazy segmentation algorithm.This article proposed algorithm, which originates from SRB algorithm however is further based on user behavior analysis, improves caching performance with a threshold introduced. In addition, we take into account popularities of video segments rather than merely entire videos. The experimental results demonstrate good performance of SRB algorithm on data of VOD system in University of Sicence and Technology of China, and more importantly the efficacy of our improvement on SRB algorithm.