A File Cache Mechanism Considering Access Cost and Being Suitable for Many-core CPU Environment

本研究旨在开发一种适用于多核CPU环境的文件缓存算法。新方法尝试缓存高访问成本的数据，如随机访问的数据，需要很长时间从二级存储设备读取。TPC-H （DBMS事实上的标准基准）的评估显示，在真实的计算机上，性能提高了大约5%。然而，几十倍或几倍的性能改进不足以实现实际使用。

This research aims to develop a novel file cache algorithm suitable for many-core CPU environment that has many computation cores. The new method tries to cache high access cost data such as random accessed data that requires long time to be read from secondary storage devices. The evaluations by TPC-H, a de facto standard benchmark for DBMS, showed about five percent performance improvement on real computers. However, several dozen or several fold performance improvement enough for actual use was not achieved.