Providing redundancy in a storage system

• US 7,996,608 B1
• Filed: 10/20/2006
• Issued: 08/09/2011
• Est. Priority Date: 10/20/2005
• Status: Active Grant

• Current Assignee: AmZetta Technologies, LLC (American Megatrends International, LLC)
• Original Assignee: American Megatrends Incorporated
• Inventors: Chatterjee, Paresh, Balakrishnan, Narayanan, Narayanan, Ajit, Muthirisavenugopal, Vijayarankan
• Primary Examiner(s): Ellis; Kevin L
• Assistant Examiner(s): Cygiel; Gary W
• Application Number:
  US11/551,291
• Time in Patent Office: 1,754 Days
• Field of Search: None
• US Class Current: 711/114
• CPC Class Codes: G06F 11/1096 Parity calculation or recalculation after configuration or reconfiguration of the system
• CPC Class Codes: G06F 11/2069 Management of state, configuration or failover

Abstract

A system, method, apparatus, and computer-readable medium are described for providing redundancy in a storage system. According to one method, maps are generated and stored that define stripe patterns for storing data on the storage nodes of a storage cluster. The maps are defined such that when a new storage node is added to the cluster, no movement of data occurs between two storage nodes that existed in the cluster prior to the addition of the new storage node during re-striping, and such that the data stored on each storage node is mirrored on another storage node. Storage nodes may also be designated as an owner or a peer for each storage zone. Input/output operations received at an owner node are fielded directly and mirrored to the peer node, while input/output operations received at a peer node are redirected to the owner node for processing.

Claims

1. A method for providing redundancy in a storage cluster comprising two or more storage computers, the method comprising:

storing a plurality of pre-defined maps on each of the two or more storage computers in the storage cluster, each pre-defined map defining a stripe pattern for storing data on a storage cluster consisting of a specific number of storage computers according to storage zones, wherein the stripe pattern for each pre-defined map is defined such that when the storage cluster is striped according to a first of the pre-defined maps, and a new storage computer is added to the storage cluster, data may be re-striped between the storage computers according to a second of the pre-defined maps such that storage zones are never moved between two storage computers that existed in the storage cluster prior to the addition of the new storage computer, regardless of the number of storage computers that exist in the storage cluster, wherein the stripe pattern for each pre-defined map is further defined such that each storage zone is stored on at least two storage computers in the storage cluster, and wherein the stripe pattern for each pre-defined map designates for each storage zone one of the at least two storage computers as an owner of the storage zone and the others of the at least two storage computers as peers for the storage zone;

striping data across the storage computers of the storage cluster according to a first map of the pre-defined maps corresponding to the number of storage computers in the storage cluster;

receiving an input/output operation directed toward a destination storage zone at a storage computer designated by the first map as the owner of the destination storage zone;

performing the input/output operation at the storage computer designated by the first map as the owner of the destination storage zone; and

mirroring the input/output operation to the storage computers designated by the first map as peers for the destination storage zone.

2. The method of claim 1, further comprising:

detecting that a new storage computer has been added to the storage cluster;

in response to detecting that the new storage computer has been added, identifying one or more storage zones to be moved to the new storage computer based on the first map and a second map of the pre-defined maps corresponding to the number of storage computers in the storage cluster after the addition of the new storage computer; and

individually moving each of the identified storage zones to the new storage computer, wherein the stripe patterns of the first map and the second map are further defined such that the designation of each storage zone to be moved to the new storage computer as an owner or a peer is maintained following the movement of the storage zone to the new storage computer.

3. The method of claim 1, wherein the input/output operation is received at a storage computer designated by the first map as a peer for the destination storage zone, and wherein the method further comprises in response thereto:

redirecting based on the first map the input/output operation to the storage computer designated by the first map as the owner of the destination storage zone;

performing the input/output operation at the storage computer designated by the first map as the owner of the destination storage zone; and

mirroring the input/output operation to the storage computers designated by the first map as peers for the destination storage zone.

4. The method of claim 1, wherein the input/output operation is received at a storage computer designated by the first map as a peer for the destination storage zone, and wherein the method further comprises in response thereto:

determining that the storage computer designated by the first map as the owner of the destination storage zone is unavailable;

in response to detecting that the storage computer designated by the first map as the owner of the destination storage zone is unavailable, performing the input/output operation at the storage computer designated by the first map as the peer for the destination storage zone; and

rebuilding the storage zone on the storage computer designated by the first map as the owner of the storage zone from the peer when the owner becomes available.

5. The method of claim 1, wherein the input/output operation is received at a storage computer not designated by the first map as either the owner or the peer for the destination storage zone, and wherein the method further comprises in response thereto:

redirecting based on the first map the input/output operation to the storage computer designated by the first map as the owner of the destination storage zone;

performing the input/output operation at the storage computer designated by the first map as the owner of the destination storage zone; and

mirroring the input/output operation to the storage computers designated by the first map as peers for the destination storage zone.

6. The method of claim 1, wherein the input/output operation comprises a read operation, and further comprising in response to receiving an input/output operation directed to a storage zone at a storage computer designated by the first map as the owner of the destination storage zone:

detecting a medium error at the storage computer designated by the first map as the owner of the destination storage zone;

in response to detecting the medium error, performing the read operation at the storage computer designated by the first map as the peer for the destination storage zone;

writing the read data at the storage computer designated by the first map as the owner of the destination storage zone; and

responding to the input/output operation with the read data.

7. A computer-readable storage medium having computer-executable instructions stored thereon which, when executed by a computer, will cause the computer to perform a method for providing redundancy in a storage cluster comprising two or more storage computers, the method comprising:

storing a plurality of pre-defined maps on each of the two or more storage computers in the storage cluster, each pre-defined map defining a stripe pattern for storing data on a storage cluster consisting of a specific number of storage computers according to storage zones, wherein the stripe pattern for each pre-defined map is defined such that when the storage cluster is striped according to a first of the pre-defined maps, and a new storage computer is added to the storage cluster, data may be re-striped between the storage computers according to a second of the pre-defined maps such that storage zones are never moved between two storage computers that existed in the storage cluster prior to the addition of the new storage computer, regardless of the number of storage computers that exist in the storage cluster, wherein the stripe pattern for each pre-defined map is further defined such that each storage zone is stored on at least two storage computers in the storage cluster, and wherein the stripe pattern for each pre-defined map designates for each storage zone one of the at least two storage computers as an owner of the storage zone and the others of the at least two storage computers as peers for the storage zone;

striping data across the storage computers of the storage cluster according to a first map of the pre-defined maps corresponding to the number of storage computers in the storage cluster;

receiving an input/output operation directed toward a destination storage zone at a storage computer designated by the first map as the owner of the destination storage zone;

performing the input/output operation at the storage computer designated by the first map as the owner of the destination storage zone; and

mirroring the input/output operation to the storage computers designated by the first map as peers for the destination storage zone.

8. A computer-readable storage medium having stored thereon:

a plurality of pre-defined maps, each pre-defined map defining a stripe pattern for storing data on a storage cluster consisting of a specific number of storage computers; and

computer-executable instructions that, when executed by storage computers in a storage cluster comprising N storage computers, N being greater than or equal to two, cause the storage computers to

stripe data across the storage computers of the storage cluster according to a first map from the plurality of pre-defined maps defining a stripe pattern for storing data on a storage cluster consisting of N storage computers;

add a new storage computer to the storage cluster; and

in response to the addition of the new storage computer to the storage cluster, re-stripe the data across the storage computers according to a second map from the plurality of pre-defined maps defining a stripe pattern for storing data on a storage cluster consisting of N+1 storage computers, wherein the stripe patterns of the first map and the second map are defined such that, during re-striping, movement of data never occurs between two storage computers that existed in the storage cluster prior to the addition of the new storage computer, regardless of the number of storage computers that exist in the storage cluster,

wherein the storage capacity of the storage cluster is divided into storage zones, the stripe pattern for each pre-defined map is further defined such that each storage zone is stored on at least two storage computers in the storage cluster, and the stripe pattern for each pre-defined map designates for each storage zone one of the at least two storage computers as an owner of the storage zone and the others of the at least two storage computers as peers for the storage zone.

9. The computer-readable storage medium of claim 8, having stored thereon further computer-executable instructions that cause the storage computers to:

receive an input/output operation at a storage computer designated by the second map as a peer node for a destination storage zone;

redirect based on the second map the input/output operation to the storage computer designated by the second map as the owner node for the destination storage zone;

perform the input/output operation at the storage computer designated by the second map as the owner node for the destination storage zone; and

mirror the input/output operation to the storage computers designated by the second map as peer nodes for the destination storage zone.

10. The computer-readable storage medium of claim 8, having stored thereon further computer-executable instructions that cause the storage computers to:

receive an input/output operation at a storage computer designated by the second map as an owner node for a destination storage zone;

perform the input/output operation at the storage computer designated by the second map as the owner node for the destination storage zone; and

mirror the input/output operation to the storage computers designated by the second map as peer nodes for the destination storage zone.