摘要
为提高RAID系统的可靠性,RAID3/5/6算法在数据写入的过程中采用异或运算产生奇偶校验信息。当RAID出现故障时,也是通过异或运算完成数据的重构。因此,异或运算是RAID系统工作时频繁而且重要的操作之一。本文详细介绍了采用Intel IOP321处理器的应用加速单元实现异或运算的工作原理和软件模块设计,并通过实验测试证明,专门的硬异或单元比嵌入式处理器做软件异或运算的速度快7倍以上,有效地解决了嵌入式环境下异或的性能瓶颈问题。
To improve the reliability of RAID storage systems,the XOR operation is performed to generate the parity check information in RAID level 3/5/6 when writing IO data,and is also used to rebuild an RAID system when the system works in the degrade mode.So the XOR operation is one of the most frequent and important operations for RAID systems.In this paper,it discusses the theory and implementation of hardware and software for the application accelerator unit in the Intel IOP321 processor,then proves by experiments that the XOR speed of the special hardware unit is 7 times faster at least than that of embedded CPUs,which can resolve the bottleneck problem of the XOR operation in embedded environments.
出处
《计算机工程与科学》
CSCD
2007年第3期87-90,共4页
Computer Engineering & Science
基金
国家自然科学基金资助项目(60173043)