基于访问预测的P2PVoD系统服务端数据缓存策略

A cache management scheme for P2P VoD systems＇ service ends based on data access forecast

分析了P2PVoD系统中各服务端数据访问顺序性与持续性的特点，提出了一种适用于P2PVoD系统的基于访问预测的服务端数据缓存策略（CAF）。该策略对顺序访问与随机访问进行预测并实施数据预取，在计算P2PVoD网络中节点数据访问持续性的基础上，由历史访问与预测的未来访问推断缓存项未来被访问频率，以该频率选择缓存替换目标。VoD系统访问记录驱动的仿真实验证明，该缓存策略可有效提高缓存命中率、降低磁盘I／O次数。

Based on the analysis of the characteristics on the continuity and sequence of the data access pattern on the service ends of a P2P VoD system, this paper proposes a cache management scheme based on access forecast for the service ends, called the CAF. This scheme predicts sequential and random accesses in the future and implements data prefetching accordingly. To deter,nine the cache eviction target, this scheme calculates the future access frequency of cache items base on historical and predicted access on the basis of getting the continuous state of the whole P2P VoD system. The simulation, driven by a VoD system＇s access records, testifies that this scheme can improve the cache hit ratio and reduce the count of HDD＇ s I/O operation effectively.