Method and apparatus for reducing partial write latency in redundant disk arrays

Claim 1

In a computer system including a processor, random access memory and a data storage system having multiple disks rotated by disk spindles in which rotational speed and angular position of said disk spindles are synchronized, said data storage system comprising: at least three disk drives, each of said disk drives having corresponding tracks and sectors for storing digital data within said sectors; control means for writing digital data to and reading digital data from said tracks and sectors of said disk drives, said control means connected to said disk drives; wherein at least one of said disk drives is used to store parity data, and the other of said disk drives is used to store general data, said parity data calculated from the general data stored on said corresponding sectors of said disk drives containing said general data; and wherein said sectors containing said parity data being position skewed in relation to said corresponding sectors of said disk drives containing said general data so that said control means may read or write said general data across corresponding data stripes and calculate parity for said corresponding data stripes of said disk drives containing said general data before writing the calculated parity onto corresponding sectors of said disk drives containing parity data and before said disk drives containing said parity data have completed a full revolution after reading or writing said general data.

Claim 8

A method for reducing write a write latency in redundant disk arrays in a computer system, said computer system including a processor, a random access memory and a multiple disk data storage system having multiple disks rotated by disk spindles, and in which rotational speed and angular position of the disk spindles are synchronized and wherein at least one of said multiple disks is a parity disk, and the other of said multiple disks are data disks, wherein parity of a data stripe write operation is calculated and written to said parity disk within a same disk rotation as a read and a write operation to said data disks, said method comprising the steps of: writing new data to sectors of at least one of said data disks; reading data from sectors of the other of said data disks which correspond with said written new data sectors; calculating parity of data stripes wherein said data stripes are comprised of said new data written and said data read; and writing said calculated parity to sectors of said parity disk which correspond with said sectors of said new data written and said data read within the same disk revolution that said new data was written, said parity disk having position skewed corresponding sectors so as to allow adequate time for reading data sectors containing existing stripe data, writing new data to corresponding data sectors, and calculating and writing parity of data stripes to said parity disk.

Claim 9

A method for reducing write a write latency in redundant disk arrays in a computer system, said computer system including a processor, a random access memory and a multiple disk data storage system having multiple disks rotated by disk spindles, and in which rotational speed and angular position of the disk spindles are synchronized and wherein at least one of said multiple disks is a parity disk, and the other of said multiple disks are data disks, wherein parity of a data stripe write operation is calculated and written to said parity disk within a same disk rotation as a read and a write operation to said data disks, said method comprising the steps of: writing new data to sectors of at least one of said data disks; reading data from sectors of the other of said data disks which correspond with said written new data sectors; calculating parity of data stripes wherein said data stripes are comprised of said new data written and said data read; and writing said calculated parity to sectors of said parity disk which correspond with said sectors of said new data written and said data read within the same disk revolution that said new data was written, said parity disk having position skewed corresponding sectors so as to allow adequate time for reading data sectors containing existing stripe data, writing new data to corresponding data sectors, and calculating and writing parity of data stripes to said parity disk.