SYSTEM AND METHOD FOR OPTIMIZING DATA MIGRATION IN A PARTITIONED DATABASE

• US 20120254175A1
• Filed: 04/01/2011
• Published: 10/04/2012
• Est. Priority Date: 04/01/2011
• Status: Active Grant

• Current Assignee: MongoDB, Inc.
• Original Assignee: MongoDB, Inc.
• Inventors: Horowitz, Eliot, Merriman, Dwight
• Granted Patent: US 9,740,762 B2
• Time in Patent Office: Days
• US Class Current: 707/737
• CPC Class Codes: G06F 16/278 Data partitioning, e.g. horizontal or vertical partitioning

Abstract

According to one aspect, provided is a horizontally scaled database architecture. Partition a database enables efficient distribution of data across a number of systems reducing processing costs associated with multiple machines. According to some aspects, the partitioned database can be manages as a single source interface to handle client requests. Further, it is realized that by identifying and testing key properties, horizontal scaling architectures can be implemented and operated with minimal overhead. In one embodiment, databases can be partitioned in an order preserving manner such that the overhead associated with moving the data for a given partition can be minimized during management of the data and/or database. In one embodiment, splits and migrations operations prioritize zero cost partitions, thereby, reducing computational burden associated with managing a partitioned database.

Claims

1. A system for optimizing data distribution, the system comprising:

at least one processor operatively connected to a memory for executing system components;

a database comprising a plurality of database partitions, wherein at least one of the plurality of database partitions includes a contiguous range of data from the database; and

a partition component configured to;

detect a partition size for the at least one of the plurality of database partitions that exceeds a size threshold;

split the at least one of the database partitions into at least a first and a second partition;

control a distribution of data within the first and the second partition based on a value for a database key associated with the data in the at least one of the plurality of database partitions, wherein controlling the distribution includes minimizing any data distributed to the second partition based on a maximum value for the database key associated with the data in the at least one of the plurality of database partitions.

2. The system according to claim 1, wherein the partition component is further configured to:

assign at least any data in the at least one of the plurality of database partitions having associated database key values less than the maximum value to the first partition; and

assign at least any data in the at least one of the plurality of database partitions having database key values greater that the maximum value to the second partition.

3. The system according to claim 1, wherein the partition component is further configured to:

identify database partitions having a sequential database key.

4. The system according to claim 1, wherein the system further comprises a plurality of servers, wherein the plurality of servers are configured to host the plurality of database partitions.

5. The system according to claim 1, further comprising a migration component configured to migrate database partitions between a plurality of servers configured to host the database partitions.

6. The system according to claim 7, further comprising a rebalancing component configured to determine a state of the database based on a distribution of the plurality of partitions across the plurality of servers, wherein the rebalancing component is further configured to migrate at least one partition in response to the state indicating an imbalanced distribution of partitions.

7. The system according to claim 4, further comprising a routing component configured to route database requests to identified partitions, wherein the routing component is further configured to identify partitions based, at least, on key values associated with the data request.

8. The system according to claim 4, further comprising a configuration component configured to manage metadata information associated with each of the plurality of partitions, the metadata information including a defined range of key values associated with each partition.

9. The system according to claim 7, wherein the configuration component is further configured to replicate the metadata across any routing component of the system.

10. The system according to claim 8, wherein the configuration component is further configured to update the metadata information in response to the partition component splitting the at least one of the plurality of database partitions into at least the first and the second partition.

11. The system according to claim 8, wherein the configuration component is further configured to update the metadata information in response to migration of database partitions between the plurality of servers.

12. The system according to claim 4, further comprising a reconciliation component configured to log database operations received on partitions during at least one of a migration operation and a splitting operation, wherein the reconciliation component is further configured to update at least one partition in response to the completion of a respective migration and splitting operation.

13. A computer implemented method for optimizing data distribution, the method comprising acts of:

monitoring, by a computer system, a distributed database including a plurality of database partitions, wherein at least one of the plurality of database partitions includes a contiguous range of data from the database;

detecting, by the computer system, a partition size of the at least one of the plurality of database partitions exceeds a size threshold;

splitting, by the computer system, the at least one of the plurality of database partitions into at least a first and a second partition;

controlling, by the computer system, a distribution of data within the first and the second partition based on a value for a database key associated with the data in the at least one of the plurality of database partitions, wherein controlling the distribution includes minimizing any data distributed to the second partition based on a maximum value for the database key associated with the data in the at least one of the plurality of database partitions.

14. The method according to claim 13, wherein the act of minimizing any data distributed to the second partition includes acts of:

assigning at least any data in the at least one of the plurality of database partitions having database key values less than the maximum value to the first partition; and

assigning at least any data in the at least one of the plurality of database partitions having database key values greater that the maximum value to the second partition.

15. The method according to claim 13, wherein the method further comprises hosting the distributed database on a plurality of servers, wherein the act of hosting the distributed database includes hosting the plurality of database partitions on the plurality of servers.

16. The method according to claim 13, further comprising an act of identifying a sequential database key, wherein the act of controlling is executed responsive to the act of identifying the sequential database key.

17. The method according to claim 15, further comprising an act of migrating database partitions between the plurality of servers.

18. The method according to claim 15, further comprising acts of:

determining a state of the distributed database based on a distribution of the plurality of partitions across the plurality of servers; and

migrating at least one partition in response to the state indicating an imbalanced distribution of partitions.

19. The method according to claim 15, further comprising an act of routing database requests to identified partitions, wherein the act of routing includes identifying partitions based, at least, on key values associated with the data request.

20. The method according to claim 15, further comprising an act of managing metadata information associated with each of the plurality of partitions, wherein the metadata information includes a defined range of key values associated with the data contained in each partition.

21. The method according to claim 20, further comprising an act of replicating the metadata across any routing component of the system.

22. The method according to claim 20, further comprising an act of updating the metadata information in response to splitting the at least one of the plurality of database partitions into at least the first and the second partition.

23. The method according to claim 20, further comprising an act of updating the metadata information in response to migration of database partitions between the plurality of servers.

24. The method according to claim 15, further comprising acts of:

logging database operations received during at least one of a migration operation and a splitting operation;

reconciling at least one partition after completing the at least one of the migration operation and the splitting operation, wherein the act of reconciling includes an act of executing the logged database operations.

25. A computer-readable storage medium having computer-readable instructions that, as a result of being executed by a computer, instruct the computer to perform a method for optimizing data distribution, the method comprising acts of:

monitoring a distributed database including a plurality of database partitions, wherein at least one of the plurality of database partitions includes a contiguous range of data from the database;

detecting a partition size of the at least one of the plurality of database partitions exceeds a size threshold;

splitting the at least one of the plurality of database partitions into at least a first and a second partition;

controlling a distribution of data within the first and the second partition based on a value for a database key associated with the data in the at least one of the plurality of database partitions, wherein controlling the distribution includes minimizing any data distributed to the second partition based on a maximum value for the database key associated with the data in the at least one of the plurality of database partitions.