US 5,331,646 A · Error correcting code technique for improving reliablility of a disk array

Abstract

An array of disk drives organized as a data storage system including n parity drives forming n parity chains where each parity drive maintains the parity bits of a different combination of data drives and each data drive is included in at least two parity chains, and where no two data drives are associated with the same combination of parity chains. The present ECC technique requires only n parity drives for up to 2.sup.n -n-1 data drives for a maximum total of 2.sup.n -1 disk drives. Each data drive is included in two or more parity chains and each parity chain is unique. The n parity drives are taken two at a time, then three at a time, and so on up to n at a time, each combination being unique, until all of the data drives are associated with parity chains. The ECC technique is implemented using a disk controller which is coupled to the array of disk drives and constantly maintains the parity information on the parity drives. The disk controller selects appropriate parity chains to reconstruct the data of any two data drives that fail, and also performs the reconstruction onto replacement disk drives. Data striping or distribution of the data and parity information on the disk drives further increases the performance of a disk array using the present ECC technique.

Claims13

Dependent claims

A data storage system, comprising:

a disk controller;
and
a number of disk drives coupled to said disk controller, said number being between (n(n+1)/2)+1 and 2n-1, inclusive, n being a positive integer greater than 2,
wherein n of said disk drives are parity drives and the remaining of said disk drives are data drives, each said parity drive being associated with a parity chain including a subset of less than all of said data drives wherein each byte of each said parity drive is a parity byte for corresponding bytes of said subset of data drives associated with said parity drive, and
wherein said parity chain associations between said parity drives and said data drives are determined by first taking said parity drives two at a time, then taking said parity drives more than two at a time until all of said data drives are associated with at least two of said parity chains and none of said parity chains are the same.

2 more elements

A data storage system, comprising:

a disk controller;
and
a number of disk drives coupled to said disk controller, said number being between (n(n+1)/2)+1 and 2n-1, inclusive, n being a positive integer greater than 2, each said disk drive being divided into a set of a predetermined number of consecutive stripes of memory, each said memory stripe being a predetermined size,
wherein said memory stripes of said disk drives contain n sets of parity stripes and said number of disk drives minus n sets of data stripes, each said parity stripe set comprising a number of parity stripes equal to said predetermined number each being said predetermined size, each said data stripe set comprising a number of data stripes equal to said predetermined number each being said predetermined size, each said memory stripe containing one of said data stripes or one of said parity stripes, and each said parity stripe set being associated with a parity chain including a subset of less than all of said data stripe sets wherein each byte of each said parity stripe is a parity byte for corresponding bytes of corresponding data stripes of said subset of data stripe sets associated with said parity stripe set,
wherein said parity chain associations between said parity stripe sets and said data stripe sets are determined by first taking said parity stripe sets two at a time, then taking said parity stripe sets more than two at a time until all of said data stripe sets are associated with at least two of said parity chains and none of said parity chain associations are the same, and
wherein said parity and data stripes are distributed into said memory stripes of said disk drives.

3 more elements

Dependencies are inferred from the claim text.

Error correcting code technique for improving reliablility of a disk array

Life of this patent

Error correcting code technique for improving reliablility of a disk array

Life of this patent

Abstract

Claims13