Sun Microsystems, Inc.

Adaptec, Inc.

Hitachi Computer Products (america), Inc.

Ebrahim Hashemi

John C Schell

John R Corbin

Lawrence J Lamers

James M Byrd

Martin E Schulze

Charles A Monia

Alex Umino

Andrew J Roy

Manuel Cisneros

A method and apparatus for fast initialization of storage devices. An apparatus is described including a redundant array of storage devices that comprises a logical volume of storage. The logical volume further comprises a plurality of stripes. A bit mapped vector provides state initialization information for each of the plurality of stripes in lieu of initializing each of the stripes. A storage controller initializes each of the plurality of stripes by using accesses to the redundant array while concurrently allowing access to the plurality of stripes. Initialization of a stripe can be done by processing a write access. Associated written data to the stripe and updating its corresponding redundancy effectively initializes that stripe. The bit mapped vector is updated to reflect the initialization of that stripe. In addition, while the redundant array is idle, further initialization of un-initialized stripes occurs while concurrently allowing access to the redundant array.

Method and apparatus for fast initialization of redundant arrays of storage devices

A data storage array employing block verification information to invoke initialization procedures. In one embodiment, a data storage subsystem includes a plurality of storage devices configured in an array and a storage controller coupled to the storage devices. The storage controller is configured to store data in the form of stripes where each stripe includes a plurality of data blocks stored across the plurality of storage devices. Each of the data blocks is associated with block verification information. Upon certain write operations, the storage controller is configured to initialize a given stripe in response to detecting a mismatch in the block verification information in at least one data block of the given stripe. In one particular implementation, the block verification contained in each of the data blocks includes a logical block address and/or a cyclic redundancy code.

Data storage array employing block verification information to invoke initialization procedures

Method and apparatus for striping spare storage in a data storage system. Specifically the present invention describes a method and apparatus for storing data and spare storage in a data storage system having a plurality of physical storage devices. The plurality of physical storage devices is partitioned into a plurality of slices. Logical units of storage (LUNs) are created, wherein each LUN contains data, spare, and data redundancy. Each of the plurality of slices is assigned to one of the plurality of LUNs so as to distribute storage of data, data redundancy, and spare across all of the plurality of physical storage devices. Distribution of spare storage is concentrated at the inner zone of each of the plurality of physical storage devices. The data and spare storage can be distributed uniformly or with varying granularities across all of the plurality of physical storage devices.

Method and system for striping spares in a data storage system including an array of disk drives

A method and apparatus for leveraging spare disks in a data storage system. Specifically the present invention describes a method for data recovery and redundancy in a data storage system having a plurality of disk drives. The data storage system is grouped into a plurality of arrays having data redundancy. The plurality of arrays is arranged in an optimum combination of arrays of mirrored pairs of disk drives, arrays of three disk drives in a redundant array of independent disks (RAID) configuration, and arrays of more than three disk drives in a RAID configuration to maximize performance while providing data redundancy. For every failure of one of said plurality of arrays due to a failed disk drive, a new array having data redundancy in a RAID configuration is dynamically created in said plurality of arrays. The new array is optimized for best performance and contains information from the failed disk drive.

Method and system for leveraging spares in a data storage system including a plurality of disk drives

A data storage subsystem including an array of storage devices and a storage controller is disclosed. In one embodiment, the array of storage devices stores information in multiple stripes. Each stripe may include a plurality of data blocks and redundancy information in the form of plurality of redundancy blocks. The redundancy information may be generated using an nth order generator polynomial such as a Reed Solomon code. The storage controller may be configured to perform modified read/write stripe updates by: (a) reading original data from a subset of data blocks in a target stripe; (b) reading the original redundancy information for that stripe; (c) comparing the original data with the new data to determine a data difference; (d) calculating a redundancy difference from the data difference; (e) applying the redundancy difference to the original redundancy information to obtain updated redundancy information, (f) writing the new data and updated redundancy information to the target stripe. Multiple erasure correction is also contemplated.

Storage array having multiple erasure correction and sub-stripe writing

The present invention provides methods for addressing an extended number of peripheral devices over a bus. A bus having an N-bit datapath is provided in a computer system. An extended address space is defined for a set of extended peripheral devices by assigning a bus ID to each extended peripheral device. Each bus ID includes a group ID and a group member ID. Each group ID and group member ID also indicates a priority of said associated device. One or more peripheral devices arbitrate for the bus. In this process, a peripheral device having the highest priority group ID and the highest priority group member ID among the arbitrating devices wins the arbitration. After arbitration, the winning device selects, for transferring data, a device by placing on the bus the bus IDs of the selecting and selected devices.

Methods for addressing extended number of peripheral devices over peripheral bus

An array of storage devices may be provided in which data is both striped and mirrored across the array. Data may be organized in stripes in which each stripe is divided into a plurality of stripe units. The stripe units may be mapped sequentially to consecutive storage devices in the array for each data stripe. Each data stripe is also mirrored within the array as a mirrored data stripe. Each mirrored data stripe is also divided into a plurality of stripe units. The stripe units of the mirrored stripes are distributed throughout the array according to a mapping that provides for load balancing during a reconstruction operation. According to one embodiment, stripe units for mirrored stripes are distributed according to a rotational group such that each mirrored stripe is rotated on the array by one more position than the previous mirrored stripe and wherein the rotational group is repeated as necessary. Alternatively, the mirrored stripe units may be distributed according to other permutations to improve load balancing during reconstruction of a failed device. In other embodiments, in addition to mapping mirrored stripe units to balance read operations during reconstruction, one or more spare storage devices may be striped throughout the array to improve load balancing for write operations during reconstruction.

Load balancing configuration for storage arrays employing mirroring and striping

The present invention provides a system of hardware and software that combine to form a redundant array of disk drives. This array provides a double means of redundancy that can reconstruct data from two simultaneously failed disks. The double redundancy means is implemented by two separate algorithms, DRR1 and DRR2. DRR1 takes the exclusive-or of all the commercial data and stores the result as redundancy data. DRR2 is similar to the algorithm described in Patel's article. As an additional feature, the implementation of DRR1 and DRR2 in the present invention is flexible enough to support either on-the-fly or read-modify-write level operation. Although both algorithms are in the prior art, the present invention improves on the run-time and space requirements of the algorithm. This improvement arises because of a novel relationship, discussed below, that exist between the individual T.sup.1 matrices and their column vectors that allows for a savings in computing the mapping of the commercial data to the redundancy data. This unique relationship is incorporated into the present systems hardware and software.

Growing community of inventors

Los Gatos, CA, United States of America

Ebrahim Hashemi