关联性视频点播系统中基于视频相似的缓存替换策略

Cache Replacement Strategy Based on Video Similarity in VOD System with Related Videos

关联性视频点播系统中的视频存在一定的关联性,用户会以极大的概率去观看与其当前观看视频相关联且相似度较大的视频。考虑到这一特性,针对P2P环境下的关联性视频点播系统,提出了一种基于视频相似的缓存替换策略。该策略根据视频的标题和简介等语义信息,基于空间向量模型实现视频相似度的计算,在进行缓存替换时,优先考虑替换掉同历史替换视频集合相似度最大的视频,且替换掉的视频的整体流行度尽可能小、副本数尽可能大。该缓存替换问题为一个多目标规划问题,将其转换为单目标规划,可形式化描述为0-1背包问题,基于贪心算法解决该问题。仿真实验表明,该策略在提高缓存内容命中率上是有效的。

The user in VoD system with related videos with great probability to choose the video has a lar- ger similarity and related with the current video. Considering this characteristic, a cache replacement strategy based on video similarity is proposed in P2P VoD system with related videos. According to the video semantic information such as the title and abstract, the video semantic similarity can be calculated by VSM（Vector space model）. When a peer＇s cache space is full, it will replace the videos which have a lager semantic similarity with the already replaced videos and the replaced content should have the smaller popularity and larger replications. The cache replacement problem can be described as a multi-goal optimizing problem, we transform it into a sin- gle-goal optimizing problem and then describe it as a 0--1 knapsack problem. A heuristic algorithm base on greedy algorithm is proposed to solve it. The simulation verifies the effectiveness of the scheme in promoting the hit ratio.