Accelerated RAID with rewind capability

Abstract

A method for storing data in a fault-tolerant storage subsystem having an array of failure independent data storage units, by dividing the storage area on the storage units into a logical mirror area and a logical stripe area, such that when storing data in the mirror area, duplicating the data by keeping a duplicate copy of the data on a pair of storage units, and when storing data in the stripe area, storing data as stripes of blocks, including data blocks and associated error-correction blocks.

Description

FIELD OF THE INVENTION [0001] The present invention relates to data protection in data storage devices, and in particular to data protection in disk arrays. BACKGROUND OF THE INVENTION [0002] Storage devices of various types are utilized for storing information such as in computer systems. Conventional computer systems include storage devices such as disk drives for storing information managed by an operating system file system. With decreasing costs of storage space, an increasing amount of data is stored on individual disk drives. However, in case of disk drive failure, important data can be lost. To alleviate this problem, some fault-tolerant storage devices utilize an array of redundant disk drives (RAID). [0003] In typical data storage systems including storage devices such as primary disk drives, the data stored on the primary storage devices is backed-up to secondary storage devices such as tape, from time to time. However, any change to the data on the primary storage device...

AI Technical Summary

Benefits of technology

[0011] To further enhance performance, according to the present invention, a memory cache (RAM cache) is added in front of the log cache, wherein incoming host blocks are first written to RAM cache quickly and the host is acknowledged. The host perceives a faster write cycle than is possible if the data were written to a data storage unit while the host waited for an acknowledgement. This further enhances the performance of the above hybrid RAID subsystem.

[0013] Read performance is further enhanced by pressing a data storage unit (e.g., disk drive) normally used as a spare data storage unit (“hot spare”) in the array, into temporary service in the hybrid RAID system. In a conventional RAID subsystem, any hot spare lies dormant but ready to take over if one of the data storage units in the array should fail. According to the present invention, rather than lying dormant, the hot spare can be used to replicate the data in the mirrored area of the hybrid RAID subsystem. Should any data storage unit in the array fail, this hot spare could immediately be delivered to take the place of that failed data storage unit without increasing exposure to data loss from a single data storage unit failure. However, while all the data storage units of the array are working properly, the replication of the mirror area would make the array more responsive to read requests by allowing the hot spare to supplement the mirror area.

[0014] The mirror area acts as a temporary store for the log, prior to storing the write data in its final location in the stripe area. In another version of the present invention, prior to purging the data from the mirror area, the log can be written sequentially to an archival storage medium such as tape. If a baseline backup of the entire RAID subsystem stripe is created just before the log files are archived, each successive state of the RAID subsystem can be recreated by re-executing the write requests within the archived log files. This would allow any earlier state of the stripe of the RAID subsystem to be recreated (i.e., infinite roll-back or rewind). This is beneficial in allowing recovery from e.g. user error such as accidentally erasing a file, from a virus infection, etc.

[0015] As such, the present invention provides a method and system of providing cost effective data protection with improved data read / write performance than a conventional RAID system, and also provides the capability of returning to a desired previous data state.

Problems solved by technology

While the data is en-route to a data storage unit through the RAM cache, power failure can result in data loss.

Images