Research On High Available Technologies For Disk Arrays

Redundant Arrays of Independent Disks (short for RAID) significantly improve the availability, including the reliability and performance, of storage subsystems through the par-allel technique and the redundancy mechanism. RAID has been widely deployed in modern storage systems. Along with the development of computer technologies, the amount of pro-duced data, which should also be processed, increases rapidly, thus putting forward higher requirements to the storage and processing capability. Although the rapid development of computer hardware technologies has greatly enhanced the capacity of storage systems, the availability still should be improved. Since the value of modern data far exceeds that of hard-ware and software of storage systems, either unexpected downtime or data corruption will bring a huge economic loss to the corporation. Accordingly, how to improve the availability of RAID attracts researchers to pay more attention to it. It also becomes an important and critical issue in the storage area.User I/O intensity can significantly impact the performance of the on-line RAID recon-struction due to the contention for the shared disk bandwidth. WorkOut, namely I/O Workload Outsourcing, reduces the user I/O intensity on the degraded RAID by effectively outsources all write requests and popular read requests originally targeted at the degraded RAID to a sur-rogate RAID during the reconstruction period. It speeds up the on-line reconstruction process and alleviates the user performance degradation caused by the disk failure and the recon-struction process, thus improving the reliability and performance of RAID-structured storage systems.Most existing RAID reconstruction algorithms must take a great deal of time to recon-struct the block-level unused storage space, thus affecting the availability of storage systems. JOR, a JOurnal-guided Reconstruction optimization, monitors the storage space utilization status at the block level to guide the reconstruction process, so only the failed data on the used stripes is recovered to the replacement disk. It reduces the number of physical disk ac-cesses caused by the reconstruction process, thus reducing the reconstruction time and the period of user performance degradation.Based on the data access method of RAID6 disk array during the degraded mode and the hotspot characteristic of the real I/O workload, a Self-Organized Strategy, short for SOS-RAID6, is proposed to improve the performance of RAID6. SOSRAID6 organizes the data on the failed disks to the corresponding parity locations on the first access. Then the later ac-cesses to it will be redirected to the parity locations rather than reading all the surviving disks. It reduces the amount of disk I/O per request, thus significantly improving the performance and availability of RAID6 in the degraded mode.The capacity of the write cache in the RAID5/6 controller has been unable to meet the performance requirement of applications. Unfortunately, the high cost of the non-volatile storage and the limited capacity of the backup battery restrict the rapid growth in the capacity of the write cache. So an adaptive compression algorithm, short for CWC, is proposed for the write cache in the RAID5/6 controller. CWC uses a lossless compression algorithm to compress the write data before writing it into the write cache. CWC extends the logical size of the write cache while its physical size remains unchanged, thus improving the buffer capacity of the write cache. Therefore, CWC achieves significant improvement in the performance and availability of RAID-structured storage systems, while it will not intimidate its reliability.Availability of RAID has been in-depth studied and a number of innovative methods have been put forward, which lay a good foundation for constructing the next-generation large-scale high available RAID-structured storage systems.