Reorganization of data under continuous workload

• US 9,009,112 B2
• Filed: 03/15/2010
• Issued: 04/14/2015
• Est. Priority Date: 03/15/2010
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Denuit, Bruno H. M., Talius, Tomas
• Primary Examiner(s): Pham, Hung Q
• Application Number:
  US12/723,696
• Publication Number:
  US 20110225122A1
• Time in Patent Office: 1,856 Days
• Field of Search: 707/634
• US Class Current: 707/634
• CPC Class Codes: G06F 16/178 Techniques for file synchronisation in file systems
• CPC Class Codes: G06F 16/278 Data partitioning, e.g. horizontal or vertical partitioning

Abstract

Architecture that provides the capability to automatically (e.g., dynamically) reorganize (repartition) an existing partition by dividing (splitting) or recombining (merging) logical databases. This reorganization can be performed to logical databases belonging to the same customer, and based on the partitioning of the tables in these databases. This can include not only splitting secondary replicas of a partition or merging secondary replicas of the partition, but also splitting off secondary replicas of the partition to create a new partition and merging two partitions into one partition. Moreover, these operations can occur while the logical databases are accepting workload (online).

Claims

1. A database management system comprising:

one or more computer readable storage media storing executable instructions that when executed by a processor provide;

a reorganization component configured to monitor performance parameters of a database relative to performance limits and perform dynamic adjustment of a composition of logical database units that provide online data of the database to increase or decrease the logical database units of the composition in response to changes that impact database performance, the reorganization component configured to merge multiple logical database units into a single logical database unit, the multiple logical database units being primary replicas of different partitions, the primary replicas receiving data modifications, the single logical database unit being a secondary replica, the secondary replica receiving the data modifications from the primary replicas.

2. The system of claim 1, wherein the logical database units operate online to provide continuous access to the online data during adjustment by the reorganization component.

3. The system of claim 1, wherein the reorganization component splits a logical database unit into two logical database units in response to an adverse impact on the database performance.

4. The system of claim 3, wherein the two logical database units filter received modifications to only allow relevant modifications for replication.

5. The system of claim 3, wherein the logical database unit is a replica or a partition of replicas, the partition or replica receiving modifications to data via a replication catch-up operation.

6. The system of claim 3, wherein the logical database unit is a replica or partition of replicas, the replica or partition receiving modifications to data via a replication copy operation that only scans source tables relevant to the replica or the partition.

7. The system of claim 1, wherein the reorganization component merges the multiple logical database units into the single logical database unit in response to determining merging the multiple database units would reduce overhead costs associated with managing the multiple logical database units independently below a threshold value.

8. The system of claim 1, wherein the primary replicas remain primary replicas after being merged into the single logical database unit.

9. A computer-implemented database management system comprising:

a reorganization component, at least partially implemented by a processor, configured to monitor parameters of a database that is operable with a physical media relative to performance limits of the database relative to replication and perform dynamic adjustments to a composition of replicas of the database configured to provide data to increase or decrease the replicas of the composition in response to changes that impact database performance, the reorganization component configured to;

commit a dynamic adjustment when fewer than all of the plurality of replicas affected by the dynamic adjustment acknowledge the dynamic adjustment, and

merge multiple replicas into a single replica, the multiple replicas being primary replicas of different partitions, the primary replicas receiving data modifications, the single replica being a secondary replica, the secondary replica receiving the data modifications from the primary replicas.

10. The system of claim 9, wherein the reorganization component splits a replica into smaller replicas in response to an adverse impact on the database performance, the smaller replicas each filter data modifications received from a primary replica to only allow relevant modifications.

11. The system of claim 10, wherein the replica receives the data modifications via a replication catch-up operation or a replication copy operation.

12. The system of claim 9, wherein the reorganization component merges the multiple replicas into the single replica in response to determining merging the multiple replicas into the single replica would reduce overhead costs otherwise associated with managing the multiple logical database units independently below a threshold value.

13. The system of claim 9, wherein the database is a distributed relational database that comprises a primary replica and secondary replicas distributed on multiple server machines, the reorganization component dynamically adjusting the composition by increasing or decreasing the secondary replicas in number according to the impact on database performance.

14. The system of claim 9, wherein the composition of replicas is a first partition that is dynamically adjusted into a first sub-partition and a second sub-partition to improve the database performance.

15. A computer-implemented database management method that executes on a processor and memory, comprising:

organizing a database as a partition of logical database units;

monitoring parameters associated with performance limits that impact performance of the database;

dynamically reorganizing a composition of the partition in response to database performance and while the logical database units are online, dynamically reorganizing including committing a dynamic reorganization when a replica of a portion of the database has not acknowledged the dynamic reorganization; and

merging multiple replicas into a single replica, the multiple replicas being primary replicas of different partitions, the primary replicas receiving data modifications, the single replica being a secondary replica, the secondary replica receiving the data modifications from the primary replicas.

16. The method of claim 15, further comprising splitting the partition into two new partitions of logical database units in response to an adverse impact on the database performance.

17. The method of claim 15, further comprising merging the partition with another partition to create a new partition as a result of determining merging the partition with another partition would improve the database performance below a threshold value.

18. The method of claim 15, further comprising splitting a logical database unit, which is a replica, into new replicas, the new replicas assigned to new partitions.

19. The method of claim 15, further comprising merging logical database units, which are replicas, into a new replica in the partition or a new partition.

20. The method of claim 15, further comprising filtering modifications sent from a source logical database unit to a receiving logical database unit at the source logical database unit or at the receiving logical database unit as part of replication of the modifications.