DISTRIBUTED TRANSACTION MANAGEMENT FOR DATABASE SYSTEMS WITH MULTIVERSIONING

• US 20120102006A1
• Filed: 10/20/2010
• Published: 04/26/2012
• Est. Priority Date: 10/20/2010
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Larson, Per-Ake, Diaconu, Cristian, Zwilling, Mike
• Granted Patent: US 8,356,007 B2
• Time in Patent Office: Days
• US Class Current: 707/703
• CPC Class Codes: G06F 16/1865 Transactional file systems
• CPC Class Codes: G06F 16/2322 using timestamps
• CPC Class Codes: G06F 16/2365 Ensuring data consistency and integrity
• 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 16/275 Synchronous replication
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

The subject disclosure relates to a distributed transaction management technique that ensures synchronization between participating nodes in a global or distributed database transaction. The technique leverages a commit protocol that uses local clocks at the respective participating nodes. Participants in a global transaction are configured to utilize the same commit timestamp and logical read time and can advance their respective local clocks to establish this synchronization. In one embodiment, distributed commit utilizes a modified version of two-phase commit that includes an extra phase to collect commit timestamp votes from participants. Additionally, a heartbeat mechanism can be used to establish loose synchronization between nodes. In another embodiment, a node can respond to a remote transaction request by returning a list of nodes involved in generating the result of the transaction and the types of access used by such nodes in addition to the transaction result itself.

Claims

1. A distributed database transaction management system, comprising:

a distributed transaction management component configured to enforce atomicity of a transaction operating on data at a plurality of distributed database nodes by facilitating synchronization of a commit timestamp used by the plurality of distributed database nodes in committing the transaction;

wherein the commit timestamp is synchronized without reference to a global clock globally available to the plurality of distributed database nodes.

2. The system according to claim 1, wherein the distributed transaction management component is further configured to synchronize a logical read time used by the transaction across the plurality of distributed database nodes.

3. The system according to claim 1, wherein the distributed transaction management component is further configured to communicate a heartbeat message that includes data relating to a system time, and wherein the heartbeat message facilitates synchronization between respective local clocks at the plurality of distributed database nodes at least in part by triggering advancement of the respective local clocks in conformance with the system time.

4. The system according to claim 1, wherein at least one distributed database node of the plurality of distributed database nodes operates as a coordinator for the transaction and includes the distributed transaction management component, the distributed transaction management component is configured to operate on a per-transaction basis, and multiple distributed database nodes are configured to simultaneously manage respective transactions via respective distributed transaction management components.

5. The system according to claim 1, wherein at least one distributed database node of the plurality of distributed database nodes is configured to manage a grouping of transactions associated with the plurality of distributed database nodes via the distributed transaction management component.

6. The system according to claim 1, wherein the distributed transaction management component comprises a commit timestamp synchronizer component configured to collect commit time votes from the plurality of distributed database nodes, to facilitate synchronization of the commit timestamp used by the plurality of distributed database nodes in committing the transaction based at least in part on the commit time votes, and to facilitate advancement of local clocks respectively associated with at least one distributed database node of the plurality of distributed database nodes in conformance with the commit timestamp.

7. The system according to claim 1, wherein the transaction is committed at the plurality of distributed database nodes based at least in part on a two-phase commit procedure.

8. The system according to claim 1, wherein the distributed transaction management component is associated with a global transaction coordinator.

9. The system according to claim 8, wherein the global transaction coordinator manages all transactions associated with the plurality of distributed database nodes.

10. The system according to claim 8, wherein the global transaction coordinator is configured to incorporate system time information into respective messages communicated to respective distributed database nodes of the plurality of distributed database nodes, and wherein the system time information facilitates synchronization of local clocks associated with the respective distributed database nodes of the plurality of distributed database nodes.

11. The system according to claim 8, wherein the global transaction coordinator is configured to incorporate active transaction information into respective messages communicated to respective distributed database nodes of the plurality of distributed database nodes, and wherein the active transaction information facilitates control of garbage collection performed by the respective distributed database nodes of the plurality of distributed database nodes.

12. A method for managing a distributed database transaction, comprising:

receiving information relating to a transaction operating on data in data stores associated with respective participating nodes associated with the transaction;

requesting commit time votes from the respective participating nodes;

receiving the commit time votes from the respective participating nodes in response to the requesting;

computing a global commit timestamp for the transaction based at least in part on the commit time votes; and

synchronizing commitment of the transaction at the respective participating nodes to the global commit timestamp.

13. The method of claim 12, further comprising:

obtaining local clock values corresponding to one or more nodes of the respective participating nodes;

generating system time information based at least in part on the local clock values; and

communicating the system time information to the respective participating nodes.

14. The method of claim 12, further comprising:

receiving begin timestamps of respective oldest active transactions of the respective participating nodes;

identifying a globally oldest active transaction from among the respective participating nodes; and

communicating a begin timestamp of the globally oldest active transaction to the respective participating nodes.

15. The method of claim 12, wherein the synchronizing comprises synchronizing commitment of the transaction at the respective participating nodes according to a two-phase commit procedure, a three-phase commit procedure or a Paxos protocol.

16. A method for managing a distributed database transaction, comprising:

receiving information relating to a distributed database transaction from a transaction coordinator associated with the distributed database transaction;

determining a commit time vote for the distributed database transaction based at least in part on a local clock;

communicating the commit time vote for the distributed database transaction to the transaction coordinator;

receiving a global commit timestamp from the transaction coordinator; and

synchronizing commitment of the distributed database transaction to the global commit timestamp.

17. The method of claim 16, further comprising:

sending a retry request for the global transaction to the transaction coordinator if the global commit timestamp indicates a time that is prior to a time associated with the local clock.

18. The method of claim 16, further comprising:

identifying system time information within one or more messages received from the transaction coordinator; and

advancing the local clock according to the system time information.

19. The method of claim 16, further comprising:

receiving a begin timestamp of an oldest active transaction from the transaction coordinator; and

configuring garbage collection such that the garbage collection does not proceed beyond the begin timestamp of the oldest active transaction.

20. The method of claim 16, further comprising:

receiving a remote request from a requesting database node corresponding to the distributed database transaction;

creating a local transaction corresponding to the distributed database transaction;

compiling a list of database nodes involved in generating a result of the local transaction and access types utilized by respective database nodes in the list of database nodes; and

returning the list of database nodes and the access types to the requesting database node in response to the remote request.