Decentralized transaction commit protocol

• US 10,235,440 B2
• Filed: 12/21/2015
• Issued: 03/19/2019
• Est. Priority Date: 12/21/2015
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP SE
• Inventors: Lee, Juchang, Park, Chang Gyoo, Kim, Kyu Hwan
• Primary Examiner(s): Oberly, Van H
• Application Number:
  US14/977,427
• Publication Number:
  US 20170177698A1
• Time in Patent Office: 1,184 Days
• CPC Class Codes: G06F 16/2315 Optimistic concurrency control
• CPC Class Codes: G06F 16/2455 Query execution
• CPC Class Codes: G06F 16/275 Synchronous replication

Abstract

Technologies are described for facilitating transaction processing within a database environment having first, second, and third database system nodes. In the database system, the first database system node receives from the second database system node a request to precommit a first database transaction. The first database system node stores information for the first database transaction that includes an indication that the second database system node coordinates the committing of the first database transaction. The first database system node receives from the third database system node a request to precommit a second database transaction. The first database system node stores information for the second database transaction that includes an indication that the third database system node coordinates the committing of the second database transaction.

Claims

1. A method, implemented at least in part by a first database system node comprising a processing unit and memory, the first database system node being in communication with at least a second and a third database system nodes, for facilitating database transaction processing within a database environment, the method comprising, at the first database system node:

receiving a request to commit a first database transaction;

sending a request to the second database system node to precommit the first database transaction;

determining a synchronized transaction token, the synchronized transaction token representing a most recent commit state known to at least the first database system node and the second database system node;

assigning, based at least in part on the synchronized transaction token, a first transaction token to the first database transaction;

sending the first transaction token to the second database system node;

committing the first database transaction;

acknowledging the commit of the first database transaction to a database client;

receiving a request from the third database system node to precommit a second database transaction; and

adding, to a precommit log stored in persistent storage at the first database system node, a precommit log entry for the second database transaction, the precommit log entry for the second database transaction comprising an indication that the third database system node coordinates the commit of the second database transaction;

wherein the first database system node acts as a coordinator node for the first database transaction and as a worker node for the second database transaction.

2. The method of claim 1, wherein each of the synchronized transaction token and the first transaction token comprise a commit ID.

3. The method of claim 1, further comprising, at the first database system node:

receiving, from the third database system node, a second transaction token associated with the second database transaction, wherein the second transaction token comprises a timestamp;

determining, based at least in part on the second transaction token, a third transaction token, wherein the third transaction token comprises a timestamp;

assigning the third transaction token to the second database transaction; and

marking the second database transaction as committed.

4. The method of claim 3, wherein determining a third transaction token comprises:

comparing the second transaction token to a local transaction token maintained by the first database system node; and

selecting as, the third transaction token, the more current value of the second transaction token and the local transaction token maintained by the first database system node.

5. The method of claim 3, further comprising receiving a request from the third database system node to commit the second database transaction.

6. The method of claim 3, wherein the third transaction token has the same value as the second transaction token.

7. The method of claim 3, wherein the commit of the second database transaction is acknowledged to a database client by the third database system node.

8. The method of claim 1, wherein the determining a synchronized transaction token comprises:

receiving a local commit timestamp from each of the at least a second and a third database system nodes;

determining a most current value among a local commit timestamp of the first database system node and the respective local commit timestamps from the at least a second and a third database system nodes; and

setting the synchronized transaction token to be the most current value.

9. The method of claim 1, further comprising:

incrementing the synchronized transaction token; and

setting the first transaction token to be the incremented synchronized transaction token.

10. The method of claim 1, wherein the first database system node maintains a first in-doubt transaction table for transactions coordinated by the first database system node, and wherein each of the at least a second and a third database system nodes maintains its own in-doubt transaction table for transactions coordinated by that database system node.

11. The method of claim 1, further comprising, without waiting to receive a precommit acknowledgement from the second database system node in response to the sending the request to the second database system node to precommit the first database transaction, starting a process to write a commit log for the first database transaction.

12. The method of claim 1, further comprising:

receiving a multi-node database statement, the multi-node database statement executable at least in part at the first database system node;

determining that the multi-node database statement is associated with an indicator that the multi-node database statement can be executed with relaxed consistency; and

based on the determining that the multi-node database statement is associated with an indicator that the multi-node database statement can be executed with relaxed consistency, without waiting to synchronize a local transaction token maintained by the first database system node with one or more transaction tokens maintained by one or more other database system nodes, or to update an interval counter or flag indicative of such synchronizing, executing the multi-node database statement, wherein the local transaction token indicates data versions visible during the execution of the multi-node database statement.

13. The method of claim 1, further comprising, at the first database system node:

experiencing a failure;

responsive to experiencing the failure, accessing the stored precommit log entry for the second database transaction to identify that the third database system node coordinated the commit of the second database transaction; and

communicating with the third database system node to determine a final committance of the second database transaction.

14. A server comprising a processing unit and memory, wherein the server is configured to operate a first database system node of a database system, the first database system node being in communication with one or more other database system nodes, wherein the server is further configured to perform database transaction processing comprising:

coordinating, as the first database system node, a commit process for a first database transaction according to a transaction commit protocol, wherein the first database system node acts as coordinator node and each of the one or more other database system nodes involved in the commit act as worker node during the commit process for the first database transaction, the coordinating comprising;

sending a request to at least a second database system node to precommit a first database transaction; and

determining a synchronized transaction token, the synchronized transaction token representing a most recent commit state known to at least the first database system node and the second database system node; and

participating, as the first database system node, in a commit process for a second database transaction according to the transaction commit protocol, wherein the first database system node acts as worker node and one of the one or more other database system nodes acts as coordinator node during the commit process for the second database transaction, the participating comprising;

receiving a request from a third database system node, which can be the second database system node, to precommit a second transaction;

adding, to a precommit log, a precommit log entry that comprises an indication that the third database system node coordinates the commit of the second database transaction;

precommitting the second transaction; and

sending a current value of a location transaction token maintained by the first database system node to the third database system node.

15. The server of claim 14, wherein the coordinating the commit process for the first database transaction includes:

receiving a request to commit the first database transaction;

to each of the one or more other database system nodes involved in the commit, sending a request to precommit the first database transaction;

assigning, based at least in part of the synchronized transaction token, a first transaction token to the first database transaction;

to each of the one or more other database system nodes involved in the commit, sending the first transaction token;

committing the first database transaction; and

acknowledging the commit of the first database transaction to a database client.

16. The server of claim 15, wherein the first transaction token is the incremented value of the synchronized transaction token.

17. The server of claim 14, wherein the participating in the commit process for the second database transaction further includes:

receiving a second transaction token from the given database system node, the second transaction token being associated with the second database transaction;

determining, based at least in part on the second transaction token, a third transaction token;

assigning the third transaction token to the second database transaction;

sending an acknowledgment to the third database system node that the third transaction token was assigned to the second database transaction;

receiving a request from the third database system node to commit the second database transaction; and

setting the second database transaction as committed.

18. The server of claim 14, wherein the first database system node maintains an in-doubt transaction table for transactions coordinated by the first database system node.

19. One or more tangible computer-readable media storing computer-executable instructions for causing a server programmed thereby to perform database transaction processing within a database environment using data manipulation language (DML) statements, the server implementing a first database system node of a database system, the first database system node being in communication with at least a second and a third database system nodes, the database transaction processing comprising:

receiving a request to precommit a first database transaction from the second database system node;

storing transaction information for the first database transaction, including adding, to a precommit log stored in persistent storage at the first database system node, a first precommit log entry that comprises an indication that the second database system node coordinates committing the first database transaction;

precommitting the first database transaction;

sending an indication of the precommitting to the second database system node;

receiving a request to precommit a second database transaction from the third database system node;

storing transaction information for the second database transaction, including adding, to the precommit log, a second precommit log entry that comprises an indication that the third database system node coordinates committing the second database transaction;

precommitting the second database transaction; and

sending an indication of the precommitting of the second database transaction to the third database system node.

20. The one or more tangible computer-readable media of claim 19, wherein the database transaction processing further comprises:

receiving, from a database client, a request to commit a third database transaction; and

coordinating, among the first database system node and the at least a second and third database system nodes, the commit of the third database transaction.

21. The one or more tangible computer-readable media of claim 20, wherein the coordinating includes adding the third database transaction to a table of in-doubt transactions maintained by the first database system node.