文章基本信息

标题：IMPROVED ALGORITHM FGF-LRU FOR I/O PERFORMANCE OF LUSTRE
本地全文：下载
作者：LI LINLIN ; SUN LIANGXU
期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2013
卷号：51
期号：2
出版社：Journal of Theoretical and Applied
摘要：In Lustre parallel file system the I/O performance of coarse-grained data access is good, while the I/O performance of fine-grained data access is relatively worse. Therefore optimizing the I/O performance of fine-grained access is the key for the improvement of the system's overall I/O performance. This study analyzed Lustre's I/O access mode, fine-grained I/O service procedure and page reclaim algorithm, and proposed a Fine-Grained First LRU (FGF-LRU) algorithm. In this algorithm, fine-grained I/O pages are retained in cache of OST and client as long as possible, to slow down the pages' sinking, and extend the time that the pages are in main memory, and thereby, decrease the number of times of disk access and reduce the disk access cost. Through comparison and analysis of the test data, the algorithm is confirmed effective; I/O performance of fine-grained access is improved without affecting the coarse-grained I/O performance, and the system overall I/O performance is improved.
关键词：Parallel file system; Lustre; Fine-grained; I/O