ENABLING EFFICIENT RECOVERY FROM MULTIPLE FAILURES TOGETHER WITH ONE LATENT ERROR IN A STORAGE ARRAY

摘要

The present invention provides a method and apparatus of managing a storage array. The method comprises: striping the storage array to form a plurality of stripes; selecting F storage chunks from each stripe as local parity chunks, and selecting another L storage chunks from the storage array as global parity chunks; performing (F+L) fault tolerant erasure coding on all data chunks in a stripe to generate (F+L) groups of parity data, and storing F groups of parity data therein into the F local parity chunks; performing cross-stripe operation on another L groups of parity data to generate L groups of global parity data, and storing them into the L global parity chunks, respectively. The apparatus corresponds to the method. With the invention, a plurality of errors in the storage array can be detected and/or recovered to improve fault tolerance and space utilization of the storage array.

主权项

1. A computer-program product for managing a storage array consisting of a number (n) storage devices, the method comprising:
one or more computer-readable storage media and program instructions stored on the one or more computer-readable storage media, the program instructions comprising: program instructions to stripe the storage array to form a number (m) stripes, wherein each stripe contains the number (n) storage chunks from the number (n) storage devices respectively, and the storage array comprises a number (n*M) storage chunks; program instructions to select a number (F) storage chunks from each stripe as local parity chunks, selecting a number (L) storage chunks from the storage array as global parity chunks, and using at least one of the storage chunks in the storage array except the local parity chunks and the global parity chunks as data chunks, wherein at least one of the storage devices in the storage array comprises one or more data chunks and one or more global parity chunks; program instructions to perform a number (F+L) fault tolerant erasure coding on data in all the data chunks in one of the number (m) stripes to generate the number (F+L) groups of parity data, and storing only the number (F) groups of parity data therein respectively into the number (F) local parity chunks of the one of the number (m) stripes; and program instructions to perform a cross-stripe operation on the number (L) groups of parity data of the number (F+L) groups of parity data to generate the number (L) groups of global parity data, and storing the number (L) groups of global parity data respectively into the number (L) global parity chunks, wherein the number (F) is less than the number (n).