SEAMLESS DATA MIGRATION ACROSS DATABASES

• US 20150254257A1
• Filed: 03/04/2014
• Published: 09/10/2015
• Est. Priority Date: 03/04/2014
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: KRITCHKO, VLADIMIR, HE, XIN, Grebenik, VLADIMIR V., YAN, KANGRONG
• Granted Patent: US 9,720,991 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/214 Database migration support
• CPC Class Codes: G06F 16/2282 Tablespace storage structures; Management thereof
• CPC Class Codes: G06F 16/2455 Query execution
• CPC Class Codes: G06F 16/27 Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
• CPC Class Codes: G06F 16/273 Asynchronous replication or reconciliation

Abstract

Techniques are disclosed for enabling the migration of data with minimized impact on consumers of the data. A data migration agent updates pointers to active data locations and coordinates a migration from a first data resource to a second data resource so that seamless migration may be carried out. A data access layer of a distributed computing environment can take advantage of the pointers and metadata written by the data migration agent to serve requests to resources of the distributed computing environment so that read availability is available regardless of migration status.

Claims

1. A method of seamless data migration, comprising:

transferring a replica of a database partition from a source partition to a target partition while maintaining read and write properties of the source partition, wherein changes in the source partition during transfer are tracked as a source delta;

disabling write operations on the source partition after the transfer of the replica is complete;

applying the source delta to the target partition upon disabling the write operations on the source partition; and

updating a pointer of a data locator with a direction to the target partition after applying the source delta to the target partition.

2. The method of claim 1, further comprising, after applying the source delta to the target partition, writing to the metadata of the source partition an indication that the source partition is inactive and writing to the metadata of the target partition an indication that the target partition is active.

3. The method of claim 1, further comprising modifying metadata of the source partition and metadata of the target partition according to a migration status.

4. The method of claim 3, wherein the migration status comprises a migration full-synchronization phase, a migration delta synchronization phase, and a switchover phase to a normal operation.

5. The method of claim 1, wherein modifying the metadata of the source partition and the metadata of the target partition at the switchover phase to the normal operation, comprises writing to the metadata of the source partition an indication that the source partition is inactive and writing to the metadata of the target partition an indication that the target partition is active after updating the pointer of the data locator with the direction to the target partition.

6. The method of claim 1, further comprising enabling write operations on the target partition after applying the source delta to the target partition.

7. The method of claim 1, further comprising:

adjusting a length of time for a migration full-synchronization according to a predefined minimum duration of time corresponding to an amount of time a data access layer caches metadata of the database to which the source partition belongs.

8. A system for a distributed database environment, comprising:

one or more storage media;

a data migration agent stored on at least one of the one or more storage media that, when executed by a processing system, manages a migration from a source to a target and adjusts pointers to active data locations according to a migration status.

9. The system of claim 8, wherein the migration status comprises a migration full-synchronization phase, a migration delta synchronization phase, and a switchover phase to a normal operation.

10. The system of claim 9, wherein the total duration of all migration phases is no shorter than a predefined minimum duration for cached data at a data access layer.

11. The system of claim 9, wherein the data migration agent, when executed by the processing system at the switchover phase, adjusts a data locator pointer from a source partition to a target partition.

12. The system of claim 9, wherein the data migration agent, when executed by the processing system at the migration full-synchronization phase, facilitates transfer of a replica of a database partition from a source partition of the source to a target partition of the target while maintaining read and write properties of the source partition and maintaining a data locator pointer to the source partition.

13. The system of claim 9, wherein the data migration agent, when executed by the processing system at the migration delta synchronization phase, disables write operations at the source partition and facilitates transfer of source delta to the target partition while maintaining read properties of the source partition and maintaining a data locator pointer to the source partition.

14. The system of claim 8, wherein the data migration agent, when executed by the processing system, manages the migration from the source to the target by:

initiating transfer of a replica of a database partition from a source partition of the source to a target partition of the target while maintaining read and write properties of the source partition;

disabling write operations on the source partition after the transfer of the replica is complete;

applying a source delta to the target partition upon disabling the write operations on the source partition;

updating source and target partition properties; and

updating a pointer of a data locator with a direction to the target partition.

15. The system of claim 14, wherein the data migration agent, when executed by the processing system, updates the source and target partition properties by writing to the metadata of the source partition an indication that the source partition is inactive and writing to the metadata of the target partition an indication that the target partition is active.

16. A method of facilitating data requests from consumers to resources in a distributed computing environment, comprising:

receiving a request for interacting with a database; and

communicating the request to the database regardless of migration state of the database.

17. The method of claim 16, wherein communicating the request to the database regardless of migration state of the database comprises:

communicating with a data locator to retrieve an indication of an active one or more partitions of the database to serve the request;

reading metadata associated with the one or more partitions to confirm that the one or more partitions indicate an active state; and

in response to the active state of the one or more partitions being confirmed, communicating the request to the database by performing the request on the one or more partitions;

in response to the active state of the identified active resource being not confirmed, communicating the request to the database by performing the request on a target resource to which at least one of the one or more partitions is being migrated.

18. The method of claim 16, wherein communicating the request to the database regardless of migration state of the database comprises:

communicating with a data locator to retrieve an indication of an active one or more partitions of the database to serve the request;

reading metadata associated with the one or more partitions to confirm that the one or more partitions indicate an active state; and

in response to the active state of the one or more partitions being confirmed, communicating the request to the database by performing the request on the one or more partitions;

in response to the active state of the identified active resource being not confirmed, communicating the request to the database by performing the request on a source resource from which at least one of the one or more partitions is being migrated.

19. The method of claim 16, wherein the request for interacting with the database is a read request, wherein communicating the request to the database regardless of migration state of the database comprises:

locating one or more active partitions of the database;

mapping the request to the one or more active partitions; and

retrieving data from the one or more active partitions.

20. The method of claim 19, wherein locating the one or more active partitions of the database comprises:

querying a data locator for an active partition, the data locator providing a directed partition;

querying metadata on the directed partition to for the active partition;

determining whether or not the data locator and the directed partition indicate a same active partition; and

in response to determining that the data locator and the directed partition indicate the same active partition;

returning the active partition;

in response to determining that the data locator and the directed partition do not indicate the same active partition;

querying metadata on a corresponding partition on a remaining database for the active partition;

selecting the active partition directed by a majority of the data locator, directed partition and the corresponding partition; and

returning the active partition.