The Implementation Of The RAID And Research Of Its High-Availability Based On Linux

The architecture of RAID controlling software has a great impact on the RAID's scalability and programming difficulty. The SCSI middle level for Linux can shield the type difference of various target drivers, and map the RAID to a block device file in the virtual file system layer, so we design the architecture of RAID controlling software based on the SCSI middle level for Linux. With the SCSI middle level, the RAID's scalability and less programming difficulty can be easily achieved. By using the architecture, the programmer can focus his attention on the designing of RAID algorithm and the implementation of RAID functions.The whole system is designed based on the hierarchy and modularization principle. As to the designing of the RAID controlling software, the RAID abstract layer is adopted, which implements the RAID's algorithm and functions. The abstract level can be seen as a virtual logical block device driver, it is a block device to its upper level, and the appliance of a block device request to the real device driver layer.The online data rebuild strategy based on the Linux OS is presented based on the research of the characteristic of the RAID's high-availability. With the strategy, the service quality can be ensured. By the analysis of the rebuilding I/O operation, design the procedure of the online data rebuild, implement the alternate operation of the normal and rebuilding I/O. The question of consistency of the online data rebuild has been solved by the analysis of the response mode of the bad disk's write request.Finally, through the performance and availability testing experiments based on various conditions, it has been proved that the system which we have designed can satisfy the expected requirement.