一种高效的磁盘队列I/O机制(英文)  

An Efficient Disk Queue I/O Mechanism

作　　者：魏青松[1] 卢显良[1] 任立勇[1] 周旭[1] 

机构地区：[1]电子科技大学计算机科学与工程学院,成都610054

出　　处：《电子科技大学学报》2003年第6期687-691,共5页Journal of University of Electronic Science and Technology of China

基　　金：信息产业部电子工业生产发展基金资助项目,编号: [2000]1277~~

摘　　要：分析了传统磁盘队列的存储管理开销和读写性能,针对磁盘队列I/O已成为影响消息服务器性能的首要瓶颈,提出了一种高效磁盘队列I/O机制—FlashQ。FlashQ采用物理上连续的磁盘块作为磁盘队列,以紧凑方式组织数据,采用延迟写策略和预先读策略提高读写性能。实验表明,FlashQ的性能比传统磁盘队列好。The storage management overload and read-write performance of traditional disk queue are analyzed. To solve the problem that disk queue I/O is the primary performance bottleneck in messaging server, an efficient disk queue I/O mechanism called FlashQ is proposed. FlashQ utilizes pre-assigned continuous disk blocks to act as disk queue and organizes data in compact layout and adopts the Lazy Write and the Ahead Read polices to elevate the performance of read-write. Experiment shows that performance of the FlashQ is much better than that of traditional disk queue.

关 键 词：消息服务器 磁盘队列 延迟写 预先读 

分 类 号：TP316.81[自动化与计算机技术—计算机软件与理论]