System and method of performing snapshot isolation in distributed databases

• US 8,935,205 B2
• Filed: 11/16/2011
• Issued: 01/13/2015
• Est. Priority Date: 11/16/2011
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP AG (SAP SE)
• Inventors: Hildenbrand, Stefan, Binnig, Carsten, Faerber, Franz, Kossmann, Donald, Lee, Juchang, May, Norman
• Primary Examiner(s): Ortiz Ditren, Belix M
• Application Number:
  US13/298,048
• Publication Number:
  US 20130124475A1
• Time in Patent Office: 1,154 Days
• Field of Search: 707/636, 707/770
• US Class Current: 707/636
• CPC Class Codes: G06F 16/2315 Optimistic concurrency control

Abstract

A system and method of performing snapshot isolation in distributed databases. Each node stores local snapshot information that enforces snapshot isolation for that node. The method includes partially processing a distributed transaction by a first node, receiving a global commit identifier from a coordinator, and continuing to process the distributed transaction, by the first node and a second node, in accordance with the global commit identifier.

Claims

1. A computer-implemented method of performing snapshot isolation in distributed databases, comprising:

implementing, on a plurality of hardware devices, a plurality of nodes that implement a distributed database;

storing, by the plurality of nodes, a plurality of local snapshot information, wherein for a particular node a corresponding local snapshot information enforces snapshot isolation for the particular node;

partially processing, by a first node of the plurality of nodes, a transaction, wherein the first node has no a-priori knowledge that the transaction is a distributed transaction;

determining, by the first node, that the transaction requires access to data from a second node and is the distributed transaction;

sending, by the first node to a hardware coordinator, a request to access a second node of the plurality of nodes;

receiving, by the first node from the hardware coordinator, a global commit identifier; and

continuing to process the distributed transaction, by the first node and the second node, in accordance with the global commit identifier, wherein continuing to process the distributed transaction includes;

checking, by the first node, that another distributed transaction has not committed;

transmitting, from the first node to the second node, the global commit identifier; and

processing the distributed transaction, by the second node, according to a snapshot of the plurality of local snapshot information on the second node that was made prior to the global commit identifier.

2. The computer-implemented method of claim 1, wherein one of the plurality of hardware devices is configured to implement at least two of the plurality of nodes.

3. The computer-implemented method of claim 1, wherein one of the plurality of hardware devices is configured to implement one of the plurality of nodes.

4. The computer-implemented method of claim 1, wherein each of the plurality of nodes stores a different part of the distributed database.

5. The computer-implemented method of claim 1, wherein the global commit identifier is one of a plurality of global commit identifiers, wherein each one of the plurality of global commit identifiers identifies a commit of each one of a plurality of distributed transactions.

6. The computer-implemented method of claim 1, further comprising:

generating, by the hardware coordinator, the global commit identifier.

7. The computer-implemented method of claim 1, further comprising:

generating, by the hardware coordinator, the global commit identifier when one of a plurality of distributed transactions commits successfully.

8. The computer-implemented method of claim 1, wherein the global commit identifier is one of a plurality of global commit identifiers, further comprising:

generating, by the hardware coordinator, the plurality of global commit identifiers when a plurality of distributed transactions commits successfully.

9. The computer-implemented method of claim 1, further comprising:

checking, by the first node, that another distributed transaction has not committed, prior to sending the request.

10. The computer-implemented method of claim 1, further comprising:

checking, by the first node, that another distributed transaction has not committed, prior to continuing to process the distributed transaction.

11. The computer-implemented method of claim 1, wherein the plurality of local snapshot information corresponds to a plurality of snapshots of the distributed database on the plurality of nodes.

12. The computer-implemented method of claim 1, wherein the plurality of local snapshot information corresponds to a plurality of snapshots of the distributed database that are created incrementally on the plurality of nodes.

13. A system for performing snapshot isolation in distributed databases, comprising:

a plurality of hardware devices that are configured to implement a plurality of nodes that implement a distributed database; and

a hardware coordinator that is configured to generate a global commit identifier,

wherein the plurality of nodes are configured to store a plurality of local snapshot information, wherein for a particular node a corresponding local snapshot information enforces snapshot isolation for the particular node,

wherein a first node of the plurality of nodes is configured to partially process a transaction, wherein the first node has no a-priori knowledge that the transaction is a distributed transaction;

wherein the first node is configured to determine that the transaction requires access to data from a second node and is the distributed transaction;

wherein the first node is configured to send, to the hardware coordinator, a request to access a second node of the plurality of nodes,

wherein the first node is configured to receive, from the hardware coordinator, the global commit identifier, and

wherein the first node and the second node are configured to continue to process the distributed transaction in accordance with the global commit identifier, wherein continuing to process the distributed transaction includes;

checking, by the first node, that another distributed transaction has not committed;

transmitting, from the first node to the second node, the global commit identifier; and

processing the distributed transaction, by the second node, according to a snapshot of the plurality of local snapshot information on the second node that was made prior to the global commit identifier.

14. The system of claim 13, wherein the hardware coordinator comprises:

a transaction controller that is configured to generate the global commit identifier; and

a transaction processor that is configured to route a local transaction to one of the plurality of nodes, and to route the distributed transaction to the transaction controller.

15. The system of claim 13, wherein the global commit identifier is one of a plurality of global commit identifiers, wherein each one of the plurality of global commit identifiers identifies a commit of each one of a plurality of distributed transactions.

16. The system of claim 13, wherein the plurality of local snapshot information corresponds to a plurality of snapshots of the distributed database on the plurality of nodes.

17. The system of claim 13, wherein the first node is configured to check that another distributed transaction has not committed;

wherein the first node is configured to transmit, to the second node, the global commit identifier; and

wherein the second node is configured to process the distributed transaction according to a snapshot of the plurality of local snapshot information on the second node that was made prior to the global commit identifier.

18. A non-transitory computer readable medium storing instructions to control a computer system for performing snapshot isolation in distributed databases, comprising:

a distributed database component that is configured to control a plurality of hardware devices to implement a distributed database on a plurality of nodes; and

a coordinator component that is configured to control a hardware coordinator to generate a global commit identifier,

wherein the distributed database component is configured to control the plurality of nodes to store a plurality of local snapshot information, wherein for a particular node a corresponding local snapshot information enforces snapshot isolation for the particular node,

wherein the distributed database component is configured to control a first node of the plurality of nodes to partially process a transaction, wherein the first node has no a-priori knowledge that the transaction is a distributed transaction,

wherein the distributed database component is configured to control the first node to determine that the transaction requires access to data from a second node and is the distributed transaction;

wherein the distributed database component is configured to control the first node to send, to the hardware coordinator, a request to access a second node of the plurality of nodes,

wherein the distributed database component is configured to control the first node to receive, from the hardware coordinator, the global commit identifier, and

wherein the distributed database component is configured to control the first node and the second node to continue to process the distributed transaction in accordance with the global commit identifier, wherein continuing to process the distributed transaction includes;

checking, by the first node, that another distributed transaction has not committed;

transmitting, from the first node to the second node, the global commit identifier; and

processing the distributed transaction, by the second node, according to a snapshot of the plurality of local snapshot information on the second node that was made prior to the global commit identifier.

19. The non-transitory computer readable medium of claim 18, wherein the coordinator component comprises:

a transaction controller component that is configured to control the hardware coordinator to generate the global commit identifier; and

a transaction processor component that is configured to control the hardware coordinator to route a local transaction to one of the plurality of nodes, and to route the distributed transaction to the transaction controller.