摘要
针对传统的文件系统(如UFS等)在支持缓存服务器时存在着元数据一致性维护、同步写操作、内存拷贝和多缓存诸多固有的缺陷,我们设计和实现了一种新的、高效的、可移植性好的文件系统Sloth。该系统在应用层实现,采用异步写操作和聚集文件等技术。仿真实验表明,Sloth文件系统有效地提高了读写磁盘的性能,大大减少了访问磁盘的次数。
Traditional file systems, such as UFSs, have the following inherent deficiencies for cache servers: meta-data coherence maintenance, sync-write operation, memory copy, and multi-buffer. We design a new, highly effective and good-portability file system. This paper introduces Sloth, which is a new file system implemented on the application layer with the technologies of async-write operation, and cluster file policy. Simulation experiments show that Sloth improves the disk I/O performance effectively and decreases the number of times of accessing the disk greatly.
出处
《计算机工程与科学》
CSCD
2003年第6期35-38,共4页
Computer Engineering & Science
基金
国家863计划资助项目(2002AAIZ2102)
