Consistency methods and systems

• US 20070214194A1
• Filed: 03/07/2006
• Published: 09/13/2007
• Est. Priority Date: 03/07/2006
• Status: Abandoned Application

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Reuter, James
• Application Number:
  US11/369,320
• Publication Number:
  US 20070214194A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/10 File systems; File servers

Abstract

Embodiments of the present invention are directed to methods for maintaining data consistency of data blocks during migration or reconfiguration of a current configuration within a distributed data-storage system to a new configuration. In one embodiment of the present invention, the current configuration is first determined to be reconfigured. The new configuration is then initialized, and data blocks are copied from the current configuration to the new configuration. Then, the configuration states maintained by component data-storage systems that store data blocks of the current and new configurations are synchronized. Finally, the current configuration is deallocated. In a second embodiment of the present invention, a current configuration is determined to be reconfigured, and, while carrying out continuing READ and WRITE operations directed to data blocks of the current configuration in a data-consistent manner, the new configuration is initialized, data blocks are copied from the current configuration to the new configuration, and the timestamp and data states for the data blocks of the current and new configurations are synchronized.

Claims

1. A method for maintaining data consistency of data blocks of a current configuration and a new configuration during migration or reconfiguration of the current configuration within a distributed data-storage system comprising component data-storage systems, the method comprising:

in a first phase, determining to reconfigure the current configuration;

in a second phase, initializing the new configuration and copying data blocks from the current configuration to the new configuration;

in a third phase, synchronizing the configuration states maintained by the component data-storage systems that store data blocks of the current and new configurations; and

in a fourth phase, deallocating the current configuration.

2. The method of claim 1 wherein the component data-storage systems of the distributed data-storage system participating in the migration or reconfiguration are within one phase of one another.

3. The method of claim 1 further including:

during the second phase, directing continuing WRITE operations to both the current and new configurations, but directing continuing READ operations to the current configuration.

4. The method of claim 1 further including:

during the third phase, directing continuing WRITE and READ operations to both the current and new configurations.

5. The method of claim 1 wherein timestamps associated with each data block in a configuration are independently managed under independent quorum-based consistency mechanisms for, and independently garbage collected for, the current and new configurations during a migration or reconfiguration from the current configuration to the new configuration.

6. The method of claim 5 wherein independently managed timestamps are not compared above a logic level managing the migration or reconfiguration.

7. Computer instructions stored within a computer-readable medium that implement the method of claim 1.

8. A distributed data-storage system comprising:

component data-storage systems;

segments of data blocks distributed across the component data-storage systems, each segment of data blocks distributed according to a configuration, during normal operation, according to two configurations, during migration, or according to two or more configurations, during reconfiguration; and

control logic within the component data-storage systems that carries out a migration or a reconfiguration operation on a segment of data blocks from a current configuration to a new configuration using synchronized, independent quorum-based consistency methods for the current and new configurations.

9. The distributed data-storage system of claim 8 wherein the control logic carries out the migration or reconfiguration operation by:

in a first phase, determining to reconfigure the current configuration;

in a second phase, initializing the new configuration and copying data blocks from the current configuration to the new configuration;

in a third phase, synchronizing the configuration states maintained by component data-storage systems that store data blocks of the current and new configurations; and

in a fourth phase, deallocating the current configuration.

10. The distributed data-storage system of claim 9 wherein the control logic carries out the migration or reconfiguration operation further by:

during the second phase, directing continuing WRITE operations to both the current and new configurations, but directing continuing READ operations to the current configuration.

11. The distributed data-storage system of claim 9 wherein the control logic carries out the migration or reconfiguration operation further by:

during the third phase, directing continuing WRITE and READ operations to both the current and new configurations.

12. The distributed data-storage system of claim 8 wherein timestamps associated with each data block in a configuration are independently managed under independent quorum-based consistency mechanisms for, and independently garbage collected for, the current and new configurations during a migration or reconfiguration from the current configuration to the new configuration.

13. The distributed data-storage system of claim 8 wherein independently managed timestamps are not compared above a logic level managing the migration or reconfiguration.

14. A method for maintaining data consistency of data blocks of a current configuration and a new configuration during migration or reconfiguration, within a distributed data-storage system, from the current configuration to the new configuration, the method comprising:

determining to reconfigure the current configuration; and

while carrying out continuing READ and WRITE operations directed to data blocks of the current configuration in a data-consistent manner,

initializing the new configuration and copying data blocks from the current configuration to the new configuration, and

synchronizing the timestamp and data states for the data blocks of the current and new configurations.

15. The method of claim 14 wherein carrying out a continuing WRITE operation directed to a data block of the current configuration in a data-consistent manner further includes:

generating a common timestamp for the WRITE operation;

directing WRITE operations corresponding to the continuing WRITE operation to both the current configuration and the new configuration using the common timestamp;

when the WRITE operations directed to both the current configuration and the new configuration complete,

returning a status, and

garbage collecting the common timestamp independently in the current and new configurations.

16. The method of claim 14 wherein carrying out a continuing READ operation directed to a data block of the current configuration in a data-consistent manner further includes:

directing READ operations corresponding to the continuing READ operation to both the current configuration and the new configuration using the common timestamp;

when the READ operations directed to both the current configuration and the new configuration complete and each returns a timestamp and data,

when the timestamps returned by the READ operations are identical, returning the data returned by one of the READ operations and a success status,

when the timestamps returned by the READ operations are not identical, but the data returned by the READ operations directed to both the current configuration and the new configuration are identical, returning the data returned by one of the READ operations and a success status, and

when neither the timestamps nor the data returned by the READ operations are identical, directing a WRITE operation to write the data returned by the READ operation directed to the current configuration to the new configuration and, when the WRITE operation succeeds, returning the data written by the WRITE operation and a success status.

17. The method of claim 14 wherein carrying out a continuing READ operation in a data-consistent manner further includes:

directing a data READ operation to one of the current and new configurations, and timestamp READ operations to both the current and new configurations;

when the READ operations directed to the current configuration and the new configuration complete,

when the timestamps returned by the READ operations are not identical, directing a READ operation to the other of the current and new configurations, and, when the data returned by both data READ operations is identical, returning the data returned by one of the data READ operations and a success status,

when the timestamps returned by the READ operations are identical, but the data returned by the READ operations directed to both the current configuration and the new configuration are identical, returning the data returned by one of the READ operations and a success status, and

when neither the timestamps nor the data returned by the two data READ operations are identical, directing a WRITE operation to write the data returned by the READ operation directed to the current configuration to the new configuration and, when the WRITE operation succeeds, returning the data written by the WRITE operation and a success status.

18. Computer instructions stored within a computer-readable medium that implement the method of claim 14.

19. A distributed data-storage system comprising:

component data-storage systems;

segments of data blocks distributed across the component data-storage systems, each segment of data blocks distributed according to a configuration, during normal operation, according to two configurations, during migration, or according to two or more configurations, during reconfiguration; and

control logic within the component data-storage systems that carries out a migration or a reconfiguration operation on a segment of data blocks from a current configuration to a new configuration using unsynchronized, independent quorum-based consistency methods for the current and new configurations.

20. The distributed data-storage system of claim 19 wherein the control logic carries out the migration or reconfiguration operation by:

determining to reconfigure the current configuration; and

while carrying out continuing READ and WRITE operations directed to data blocks of the current configuration in a data-consistent manner,

initializing the new configuration and copying data blocks from the current configuration to the new configuration, and

synchronizing the timestamp and data states for the data blocks of the current and new configurations.

21. The distributed data-storage system of claim 19 wherein carrying out a continuing WRITE operation directed to a data block of the current configuration in a data-consistent manner further includes:

generating a common timestamp for the WRITE operation;

directing WRITE operations corresponding to the continuing WRITE operation to both the current configuration and the new configuration using the common timestamp;

when the WRITE operations directed to both the current configuration and the new configuration complete,

returning a status, and

garbage collecting the common timestamp independently in the current and new configurations.

22. The distributed data-storage system of claim 19 wherein carrying out a continuing READ operation directed to a data block of the current configuration in a data-consistent manner further includes:

directing READ operations corresponding to the continuing READ operation to both the current configuration and the new configuration using the common timestamp;

when the READ operations directed to both the current configuration and the new configuration complete and each returns a timestamp and data,

when the timestamps returned by the READ operations are identical, returning the data returned by one of the READ operations and a success status,

when the timestamps returned by the READ operations are not identical, but the data returned by the READ operations directed to both the current configuration and the new configuration are identical, returning the data returned by one of the READ operations and a success status, and

when neither the timestamps nor the data returned by the READ operations are identical, directing a WRITE operation to write the data returned by the READ operation directed to the current configuration to the new configuration and, when the WRITE operation succeeds, returning the data written by the WRITE operation and a success status.

23. The distributed data-storage system of claim 19 wherein carrying out a continuing READ operation in a data-consistent manner further includes:

directing a data READ operation to one of the current and new configurations, and timestamp READ operations to both the current and new configurations;

when the READ operations directed to the current configuration and the new configuration complete,

when the timestamps returned by the READ operations are not identical, directing a READ operation to the other of the current and new configurations, and, when the data returned by both data READ operations is identical, returning the data returned by one of the data READ operations and a success status,

when the timestamps returned by the READ operations are identical, but the data returned by the READ operations directed to both the current configuration and the new configuration are identical, returning the data returned by one of the READ operations and a success status, and

when neither the timestamps nor the data returned by the two data READ operations are identical, directing a WRITE operation to write the data returned by the READ operation directed to the current configuration to the new configuration and, when the WRITE operation succeeds, returning the data written by the WRITE operation and a success status.

24. A distributed data-storage system composed of component data-storage systems across one or more of which data segments are distributed, the distributed data-storage system providing for reconfiguration of a data segment from distribution across a first set of component data-storage systems to distribution across a second set of component data-storage systems, the distributed data-storage system comprising:

the component data-storage systems;

a quorum-based consistency mechanism that maintains data consistency of a data segment distributed across a set of component data-storage systems according to a current configuration; and

a means for employing two independent quorum-based consistency mechanisms for maintaining data consistency of a data segment distributed across a first set of component data-storage systems and distributed across a second set of component data-storage systems during reconfiguration of the data segment.