Disk dual copy methods and their performance

Performance analyses of three dual-copy methods for ensuring fault tolerance of disk storage devices are considered: an algorithm using synchronized disks, and single-read and parallel-read algorithms using nonsynchronized disks. With the synchronized algorithm, dual copying of data is performed almost simultaneously for every write request. With nonsynchronized disks, the two write operations are essentially sequential. Thus the coordination of the two physical data transfers for each write request degrades performance. On the other hand, the availability of dual copies of the data makes it possible to save on rotational latency time by connecting to that device whose rotational position is closest to the read/write head, thus improving performance.< >