通过分析多用户数据请求规律以及实时分解随机请求序列来获取顺序请求序列。基于对多用户顺序请求进行命令预分解和命中率统计,实现读预取长度自我学习。分析多用户预取率及系统负载与预取失效代价之间的关系,对常规自适应Cache策略进行优化,选择合适预取阈值等参数。与常规自适应预取策略相比,动态调整Cache策略的预取命中率提高了30%。有效解决了多用户访问共享存储系统的预取失效率高问题.
Dynamic Cache prefetch algorithm is put forward in this paper. Through analysis of the multi-user data requesl rule, introduction of a dynamic adjustment Cache mechanism. Based on decomposition of multiple users request command,and recording the hit rate statistics by real-time, so to realize muhi-user access length self-learning in the shared storage system. Af- ter analyzing the simulation results, selecting the appropriate dynamic parameters has improved the prefeteh hit rate performanee 30% higher than the constant adaptive algorithm.