UGD:面向固态盘缓存系统的数据过滤技术

UGD: A novel filtering method for SSD-based cache system

为了提高固态盘(SSD)缓存系统固态盘的使用寿命,研究了系统的数据过滤技术。针对现有固态盘缓存系统数据过滤技术没有考虑数据页重用距离,导致重用距离大于固态盘缓存大小的数据页进入固态盘缓存,造成不必要的缓存页替换和固态盘擦写,从而降低缓存命中率和固态盘使用寿命的问题,提出了一种协同的数据过滤技术UGD。该技术综合考虑了数据页访问频率和数据页重用距离,让访问次数少和重用距离长的数据页不进入固态盘缓存,从而增长了固态盘的使用寿命。为了评测UGD技术的效果,实现了一个固态盘缓存模拟器,并使用开源负载和实际应用负载进行了大量实验。实验数据表明,与现有过滤技术相比,UGD技术能够使固态盘缓存系统的平均命中率提高10%,使固态盘平均写入量降低42.5%。

To increas the lifetime of the solid state disk （SSD） in a SSD-based cache system, the data filtering for the system was studied, and the study revealed that existing filtering techniques do not take the data re-use distance in- to account, which makes the data pages with the large re-use distance enter into the SSD cache, causing the non- necessary page replacement and SSD writing, then the decline in the cache hit rate and the SSD lifetime. In view of this problem, a novel data filtering method, called unified ghost cache and data cache （ UGD）, was proposed. The UGD technique concurrently considers the page access count and the re-use distance to filter out the pages of large re-use distance and small access count to increase the SSD lifetime. A cache system simulator was implemented and number of extensive experiments were conducted by using public traces and traces of real-world workloads. The ex- perimental results show that in comparison with LARC, the UGD improved the cache hit rate by 10% and reduced write amount by 42.5% on average.