Abstract: A parity which is used to a disk array is formed by using a cache memory. A storing method of the cache memory is classified into three methods such that: only data is stored; only redundant information is stored; and both of the data and the redundant information are stored. Even when a power source is shut off during the data writing process, the processing stages of the data writing and parity updating and the necessary data are stored into a non-volatile memory and are used for recovery.

Abstract: A parity which is used to a disk array is formed by using a cache memory. A storing method of the cache memory is classified into three methods such that: only data is stored; only redundant information is stored; and both of the data and the redundant information are stored. In order to save processing time, a new set of redundant information (or parity data) is not calculated for every data update. Instead, the new parity data is only calculated after there have been a predetermined number of data updates. In addition, even when a power source is shut off during the data writing process, the processing stages of the data writing and parity updating and the necessary data are stored into a non-volatile memory and are used for recovery.

Abstract: When an amount of write data instructed by an upper-level system is small, a plurality of disk units are accessed individually and data split in sectors is written therein (Level 5 RAID). When an amount of write data instructed by the upper-level system is large, the plurality of disk units are accessed in parallel and data split in sectors is written therein (Level 3 RAID). When a disk unit in an array, to which a setup instruction is issued according to a processing request sent from the upper-level system, returns a fault reply, an ID management table is used to allocate an auxiliary disk instead of the failing disk unit. After the allocation, data is restored to the replacement disk using the data in normal disk units in the same rank. In a dual-port access configuration, when two transactions access disk units connected to the same port, a deadlock may occur depending on the access procedure.

Abstract: When an amount of write data instructed by an upper-level system is small, a plurality of disk units are accessed individually and data split in sectors is written therein (Level 5 RAID). When an amount of write data instructed by the upper-level system is large, the plurality of disk units are accessed in parallel and data split in sectors is written therein (Level 3 RAID). When a disk unit in an array, to which a setup instruction is issued according to a processing request sent from the upper-level system, returns a fault reply, an ID management table is used to allocate an auxiliary disk instead of the failing disk unit. After the allocation, data is restored to the replacement disk using the data in normal disk units in the same rank. In a dual-port access configuration, when two transactions access disk units connected to the same port, a deadlock may occur depending on the access procedure.