双控制器RAID系统的研究与实现

发布时间：2018-05-07 16:57

本文选题：双控制器 + 磁盘阵列　；参考：《华中科技大学》2012年硕士论文

【摘要】：随着全球信息化建设的高速发展，大量的数据业务需要专用的存储系统提供高速、稳定的数据存储服务。而双控制器RAID系统相对于单控制器RAID系统在控制器级增加硬件冗余，使其能够在磁盘级、控制器级都具有较高的稳定性和可靠性，，能够满足对存储服务稳定性要求较高的行业。相对于单控制器RAID系统而言，双控制器RAID系统实现的难点在于故障检测、故障处理与CACHE之间数据一致性的保证。通过在两个控制器之间采用专用的心跳模块进行数据通信，包括心跳数据包、状态通知等，以便在系统运行时获取镜像控制器的工作状态，当其中一个控制器出现故障时能够及时检测到。心跳模块之间通过固定的通信协议在系统启动、故障处理、故障恢复时设置控制器的工作模式，使系统能够正常工作。对于CAHCE之间的数据一致性保证，则需要每个控制器中CACHE模块维护CACHE（缓存）和MIRROR（镜像）两种数据，并且系统在处理写请求时，采用同步写入的方式，将数据同时写入到两个控制器中才返回。当系统出现故障时，控制器中的MIRROR数据将会作为CACHE数据继续提供数据服务，避免服务中断与数据丢失；在故障恢复后，两个控制器的CACHE模块需要在确保数据一致性后再继续对外提供数据服务。最后，将双控制器RAID系统与单控制器RAID系统进行性能对比测试和双控制器RAID系统的稳定性测试。显示在性能上双控制器RAID系与单控制器RAID系统大致相同。在单个控制器出现故障时，系统能够继续提供数据服务，并且没有造成数据丢失。
[Abstract]:With the rapid development of global information construction, a large number of data services need special storage systems to provide high-speed and stable data storage services. Compared with the single controller RAID system, the dual-controller RAID system increases the hardware redundancy at the controller level, which makes it more stable and reliable at the disk level and controller level, and can meet the requirements of storage service stability. Compared with the single-controller RAID system, the difficulty in the realization of double-controller RAID system lies in the guarantee of data consistency between fault detection, fault processing and CACHE. By using a special heartbeat module between the two controllers for data communication, including heartbeat packets, status notifications, etc., in order to obtain the working state of the mirror controller while the system is running. It can be detected in time when one of the controllers fails. In order to make the system work normally, the heartbeat module sets the working mode of the controller when the system starts up, the fault is handled and the fault is restored by a fixed communication protocol. For the data consistency between CAHCE, the CACHE module in each controller is required to maintain the cache (cache) and mirroring (mirroring) data, and the system uses the method of synchronous writing when processing the write request. Data is written to both controllers before it is returned. When the system fails, the MIRROR data in the controller will continue to serve as CACHE data to avoid service disruption and data loss. The CACHE modules of the two controllers need to ensure data consistency before continuing to provide data services. Finally, the performance of dual controller RAID system is compared with that of single controller RAID system and the stability of double controller RAID system is tested. The performance of dual controller RAID system is similar to that of single controller RAID system. When a single controller fails, the system can continue to provide data services without causing data loss.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP333

【参考文献】