摘要
日志结构分布式文件系统(HDFS-based Log-structured File System,HLFS)是基于Hadoop分布式文件系统(Hadoop Distributed File System,HDFS)构建的云存储平台,为了提高其数据读写性能,结合HLFS客户端与后台存储集群相分离的特点,利用最近最久未使用算法作为缓存策略,在HLFS客户端内存中设计并实现HLFS缓存,并采用固定时间间隔刷新缓存的策略来保证数据一致性。在集群系统上对所设计的缓存进行了数据读写测试,结果表明,该设计能够有效地降低HLFS与后台集群的交互频率,减少用户感知的数据读写时间。
In order to enhance the overall I/O performance of the distributed cloud storage plat- form HLFS（HDFS-based Log-structured File System）, cache module is proposed and implemen- ted in the HLFS client RAM by making use of LRU algorism due to the characteristic of HLFS of which client is separated with the back-end storage cluster. To ensure the data consistence, the data stored in the cache is designed to be written back at fixed time intervals. The cache module is tested in a cluster by reading and writing data to HLFS. The test results shows that, by enabling the cache module, HLFS can decrease the interaction frequency with the back-end cluster, there- by improving the I/O performance effectively.
出处
《西安邮电大学学报》
2013年第5期76-80,共5页
Journal of Xi’an University of Posts and Telecommunications
基金
陕西省教育厅自然科学基金资助项目(11JK1069)
