Abstract: Based on multiple RAID controllers system，this paper presents a RAID50 massive storage model.First，build hardware RAID5s of the same type on every single RAID controller，all of which have the same disk type and disk number.Secondly，regard each RAID5 as a single disk，and combine them to be a RAID50 by using software.By adopting a two level concurrency data partioning and placing method and set extended data block，which equals to a data parity group of RAID5 in a RAID controller，as the transmission granularity between Cache and RAID system，this paper carries out the full concurrency access of the m*n disks of RAID system.The two level address mapping algorithm and the best-fit algorithm of logic volume management are designed.The theoretical analysis and test results show that the method has reduced the I/O response time of the system greatly and obtained massive logic volume capacity and I/O performance which are related to the number of channels.

摘要： 提出了一种跨多阵列通道的海量存储RAID50模型，通过采取多阵列卡的RAID0分条和阵列卡上多磁盘RAID5分条和校验的二级并发的数据组织与分块方式，以扩展块（大小等于阵列卡上的一个RAID5校验组）作为Cache和阵列之间数据交换的单位，实现了将阵列矩阵中所有磁盘的容量聚合及全并发访问。设计了该模型逻辑卷管理的最佳适配算法及二级地址映射算法。理论分析与实验结果表明：该策略将I/O响应时间降到了最低，且获得了与阵列通道数线性相关的逻辑卷容量和I/O性能。