固态盘热数据识别算法研究

Research on Hot Data Identification Algorithm for Solid State Disk

固态硬盘的技术在不断提升,在读写速度、功耗、安全性、防震抗摔性、体积等方面的优势也更加明显。其中,热数据识别技术是固态硬盘技术研究中非常关键的问题,热数据的准确识别及合理分配可以改善固态硬盘的读写性能和使用寿命。文章提出了一种基于K均值聚类和K近邻分类融合的热数据识别算法——KKH。首先,根据访问请求的大小,采用K均值算法对请求进行聚类,判断该请求访问数据的冷、热程度;其次,根据请求的逻辑页地址,采用K近邻分类算法对该请求进行分类;最后,若两种方法的分类结构不一致,根据逻辑页地址采用最近邻原则对判定结果进行修正。实验结果显示,融合的热数据识别算法可以有效改善错误识别率和内存空间开销。

The technology of solid state disk has been improved, such as the speed of read and write, power consumption, security, shock resistance, volume etc. Its advantages become more obvious. The technology of hot data identification is the key problem of SSD design. The accurate identification and reasonable distribution of hot data can improve the performance of read and write and service life of SSD. A novel hot data identification algorithm was present in this paper based on fusion of K-means and K- nearest neighbor. Firstly, according to the size of access request, it was gathered via K-means algorithm to judge the degree of heat or cold. Secondly, the request was classified via K-nearest neighbor algorithm according to request of the logical page address. Finally, if the requested judgment of K-means and K- nearest neighbor was different, judgment result was revised via the nearest neighbor principle according to the logical page address. The experimental results showed that the fusion algorithm could effectively improve the rate of false recognition and memory space overhead.