Abstract: As Video-on-Demand(VoD) users may stop watching in the middle of a movie, so different parts of the same movie have different playback probabilities, which cause potential load unbalance problem for traditional segment based storage scheme. This paper analyzes the user’s behavior pattern statistically and proposes a probability model to describe it. According to the model, it proposes an improved segment based storage scheme, and demonstrates its advantages in better load balance and system efficiency through simulation.

Key words: Video-on-Demand (VoD), clustered system, data storage, load balance

摘要： 在VoD点播中，用户主动中断点播会造成影片不同时间段的部分被访问的概率不同，给传统的分块策略带来了负载不均衡问题。该文分析用户在VoD点播时中断点播的行为模式，在此基础上建立分析点播中断概率的模型，提出一种改进的分块存储策略。仿真表明，改进的策略能够更好地实现负载均衡，提高系统运行效率。

关键词: 视频点播, 集群系统, 数据存储, 负载均衡

CLC Number: 

• TP37