Method for employing doubly striped mirroring of data and reassigning data streams scheduled to be supplied by failed disk to respective ones of remaining disks

Claim 1

A method of storing data in a plurality of N independent disks comprising the steps of dividing the storage area of each of said disks into primary and secondary sections, partitioning said data into a plurality of successive blocks of data and storing said successive blocks of data in sequence in respective ones of the primary sections of said disks in a predetermined order, storing in sequence the blocks of data that have been stored in one of said primary sections of one of said disks in respective secondary sections of the other ones of said disks, unloading in sequence a block of data from the primary section of each of the N disks during respective Nth cycles and supplying the unloaded block of data to a particular one of a plurality of data streams being served by the N disks, where N>1, and responsive to one of the N disks failing, re-assigning the data streams scheduled to be supplied by the failed one of said disks for a respective one of the cycles to respective ones of the remaining disks containing a restart data block for a respective one of such data streams, in which said re-assigning is constrained so that no more than n/N additional streams are assigned to any one of the remaining disks, where n is the maximum number of data streams that can be served concurrently by any one of the disks in one cycle.

Claim 2

A method of storing data in a plurality of N independent disks comprising the steps of dividing the storage area of each of said disks into primary and secondary sections, partitioning said data into a plurality of successive blocks of data and storing said successive blocks of data in sequence in respective ones of the primary sections of said disks in a predetermined order, storing in sequence the blocks of data that have been stored in one of said primary sections of one of said disks in respective secondary sections of the other ones of said disks, unloading in sequence a block of data from the primary section of each of the N disks during respective Nth cycles and supplying the unloaded block of data to a particular one of a plurality of data streams being served by the N disks, where N>1, and responsive to one of the N disks failing, re-assigning the data streams scheduled to be supplied by the failed one of said disks for a respective one of the cycles to respective ones of the remaining disks containing a restart data block for a respective one of such data streams, in which said re-assigning is constrained so that no more than n/N additional streams are assigned to any one of the remaining disks, where n is the maximum number of data streams that can be served concurrently by any one of the disks in one cycle.