基于语义信息的cache管理策略

A Semantic Information-based Cache Management Strategy

针对传统的cache在预取时不判断预取数据块的状态,导致一些不必要的I/O,同时降低cache命中率的缺点,提出了一种基于语义信息的cache管理策略。该策略首先通过收集语义信息让磁盘了解文件系统在磁盘上的数据布局,磁盘上每个数据块是活跃的还是死亡的,并得出磁盘上分区数据块的活跃度。然后根据语义信息在预取的时候不预取死亡的块,在活跃度高的分区上提高预取参数,而在cache替换出数据块时对于死亡块不进行写盘操作。实验结果表明该策略可以较好提高cache命中率进而提高系统的吞吐量。

When the traditional cache prefects the data blocks, it will not judge that these blocks are active or dead. So this way will cause many unnecessary I/O, and it will decrease the hit rate. For ad- dressing this important issue, in this paper, a semantic information -based cache management strategy is present. Firstly, the strategy collects the semantic information to let the disk know data placement of the file system on disk, the dead data blocks and active data blocks, and active degree. Then according to the semantic information the cache can not prefect the dead data blocks, meanwhile the cache can prefect more when the rate of active degree is high. When the data block in the cache is replace and they are dead, they will not be flushed into the disk. The experiment result shows that the strategy can improve the cache hit rate, so it improves the system throughput