Method and apparatus for transparent, real time reconstruction of

Abstract

A method and apparatus for providing transparent, real time reconstruction of corrupted data from a redundant array of storage units in a manner that imposes no performance degradation for reconstruction of such corrupted data and that does not tie up system resources. The present invention is particularly suited to applications in which data is retrieved from storage equipment in approximately equally sized parcels. Each parcel is divided into 'N' blocks of data. An 'N+1st' parity block for each parcel is derived based on the content of the 'N' blocks of data. The 'N' blocks of data and the parity block for each parcel are stored in stripes of blocks in a redundancy group consisting of 'N+1' data storage units. Data is read from the storage units an entire stripe at a time. To read a stripe, a read request is concurrently sent to each storage unit in the redundancy group, requesting the block of data in that storage unit that corresponds to the stripe being read. Each storage unit independently processes the read request it has received and sends back the requested block as soon as it has been read. As soon as the first 'N' blocks are received, they are XOR'd together to reconstruct the outstanding 'N+1st' block. The 'N' received blocks and the XOR'd sum, representing a reconstruction of the outstanding 'N+1st' block, are delivered to the requesting device or process. The same procedure is used each time data is read. The present invention thus delivers data at a consistent rate regardless of whether a storage unit failure occurs.