Fault-tolerant timestamp generation for multi-node parallel databases

• US 6,125,368 A
• Filed: 02/28/1997
• Issued: 09/26/2000
• Est. Priority Date: 02/28/1997
• Status: Expired due to Term

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle Corporation
• Inventors: Loaiza, Juan R., Lee, J. William, Bridge, William, Krishnapuram, Madhusudan
• Primary Examiner(s): Kulik, Paul V.
• Application Number:
  US08/808,582
• Time in Patent Office: 1,306 Days
• Field of Search: 707/10, 707/8, 707/201, 707/202, 370/473, 375/354
• US Class Current: 1/1
• CPC Class Codes: G06F 11/1415 at system level
• CPC Class Codes: G06F 2201/835 Timestamp
• CPC Class Codes: Y10S 707/99952 Coherency, e.g. same view to multiple users
• CPC Class Codes: Y10S 707/99953 Recoverability

Abstract

Every node in a multi-node parallel database system maintains a logical clock for generating timestamps. The logical clocks are synchronized by attaching a current timestamp to every message that is sent by a node. When a node receives an incoming timestamp that is greater than the value indicated by the associated logical clock, it sets the associated logical clock forward to at least the value of the timestamp. Each node continually sends and receives a message including a timestamp at least once in a prescribed period of time. Moreover, each node maintains a high watermark that represents a time greater or equal to the highest time of any logical clock in the multi-node parallel database system.

Claims

1. A method for maintaining logical clocks in a multi-node parallel database system that includes a plurality of nodes, wherein each node of said plurality of nodes is associated with a logical clock, the method comprising the computer-implemented steps of:

causing each node of said plurality of nodes to maintain a watermark that represents a time at least as recent as the latest time represented by the logical clocks on each of said plurality of nodes; and

when the time represented on a logical clock of a node of said plurality of nodes is within a predetermined threshold of the time represented by said watermark, then causing all nodes of said plurality of nodes to increase said watermark.

2. The method of claim 1, further comprising the steps of:

causing all nodes of said plurality of nodes to stop the logical clock associated with said node, before the step of causing all nodes of said plurality of nodes to increase said watermark; and

causing all nodes of said plurality of nodes to resume the logical clock associated with said node, after the step of causing all nodes of said plurality of nodes to increase said watermark.

3. The method of claim 1 further comprising the steps of causing each node of said plurality of nodes to perform the following steps:

transmitting a first message that contains a first message timestamp representing time indicated by the logical clock associated with said node to another node of said plurality of nodes;

receiving a second message from another node of said plurality of nodes, wherein said second message contains a second message timestamp; and

in response to receiving said second message, performing the steps of;

inspecting the logical clock associated with the node to determine a time indicated by said logical clock,

comparing the time indicated by said logical clock with a time indicated by the second message timestamp, and

if the time indicated by the second message timestamp is more recent than the time indicated by the logical clock, then setting the logical clock associated with the node to reflect a time that is at least as recent as the time reflected in the message timestamp.

4. The method of claim 1, further comprising the steps of:

detecting failure of one of said plurality of nodes; and

in response to detecting failure of said one of said plurality of nodes, causing a recovery node to perform the steps of;

setting the logical clock associated with the recovery node to reflect a time that is at least as recent as the time reflected in the watermark and

after setting the logical clock, causing said recovery node to recover the node that failed.

5. A method for maintaining logical clocks in a distributed system that includes a plurality of nodes, wherein each node of said plurality of nodes is associated with a logical clock, the method comprising the computer-implemented step of:

causing each node of said plurality of nodes to perform the following steps within each of a plurality of successive periods having a common prescribed duration;

(a) transmitting a first message that contains a first message timestamp representing time indicated by the logical clock associated with said node to another node of said plurality of nodes;

(b) receiving a second message from another node of said plurality of nodes, wherein said second message contains a second message timestamp; and

(c) causing the logical clock associated with said node to reflect a time at least as recent as the time in the second message timestamp;

wherein the step of causing the logical clock associated with said node to reflect a time at least as recent as the time in the second message timestamp includes the steps of;

inspecting the logical clock associated with the node to determine a time indicated by said logical clock,

comparing the time indicated by said logical clock with a time indicated by the second message timestamp, and

if the time indicated by the second message timestamp is more recent than the time indicated by the logical clock, then setting the logical clock associated with the node to reflect a time that is at least as recent as the time reflected in the second message timestamp.

6. The method of claim 5, further comprising the steps of:

causing a central node of said plurality of nodes to transmit a first message timestamp representing time indicated by the logical clock associated with said central node to each other node of said plurality of nodes and

causing each other node of said plurality of nodes to transmit a first message timestamp representing time indicated by the logical clock associated with said node to said central node.

7. The method of claim 5, further comprising the steps of:

causing each node of said plurality of nodes to maintain a watermark that represents a time at least as recent as the latest time represented by the logical clocks on each of said plurality of nodes;

when the time represented on a logical clock of a node of said plurality of nodes is within a predetermined threshold of the time represented by said watermark, then causing all nodes of said plurality of nodes to increase said watermark;

detecting failure of one of said plurality of nodes; and

in response to detecting failure of said one of said plurality of nodes, causing a recovery node to perform the steps of;

setting the logical clock associated with the recovery node to reflect a time that is at least as recent as the time reflected in the watermark and

after setting the logical clock, causing said recovery node to recover the node that failed.

8. The method of claim 7, further comprising the steps of:

causing all nodes of said plurality of nodes to stop the logical clock associated with said node, before the step of causing all nodes of said plurality of nodes to increase said watermark and

causing all nodes of said plurality of nodes to resume the logical clock associated with said node, after the step of causing all nodes of said plurality of nodes to increase said watermark.

9. A multi-node parallel database clock control system, comprising:

a plurality of nodes, wherein each node of said plurality of nodes is associated with a logical clock;

wherein each node of said plurality of nodes is configured to maintain a watermark that represents a time at least as recent as the latest time represented by the logical clocks on each of said plurality of nodes;

wherein a watermark node of said plurality of nodes is to configured to cause all nodes of said plurality of nodes to increase said watermark, when the time represented on a logical clock of said watermark node of said plurality of nodes is within a predetermined threshold of the time represented by said watermark.

10. The multi-node parallel database clock control system of claim 9 wherein:

each node of said plurality of nodes is configured to transmit a first message that contains a first message timestamp representing time indicated by the logical clock associated with said node to another node of said plurality of nodes and

each node of said plurality of nodes is configured to a second message from another node of said plurality of nodes, wherein said second message contains a second message timestamp, and in response to receiving said second message, to perform the steps of;

inspecting the logical clock associated with the node to determine a time indicated by said logical clock,

comparing the time indicated by said logical clock with a time indicated by the second message timestamp, and

if the time indicated by the second message timestamp is more recent than the time indicated by the logical clock, then setting the logical clock associated with the node to reflect a time that is at least as recent as the time reflected in the message timestamp.

11. The multi-node parallel database clock control system of claim 9, wherein a detecting node of said plurality of nodes is configured to detect failure of one of said plurality of nodes and, in response to detecting failure of said one of said plurality of nodes, to cause a recovery node to perform the steps of:

setting the logical clock associated with the recovery node to reflect a time that is at least as recent as the time reflected in the watermark and

after setting the logical clock, causing said recovery node to recover the node that failed.

12. A distributed clock synchronizing system comprising:

a plurality of nodes, wherein each node of said plurality of nodes is associated with a logical clock;

wherein each node of said plurality of nodes is configured to perform the following steps within each of a plurality of successive periods having a common prescribed duration;

(a) transmitting a first message that contains a first message timestamp representing time indicated by the logical clock associated with said node to another node of said plurality of nodes;

(b) receiving a second message from another node of said plurality of nodes, wherein said second message contains a second message timestamp, and

(c) causing the logical clock associated with the node to reflect a time that is at least as recent as the time reflected in the second message timestamp;

wherein the step of causing the logical clock associated with said node to reflect a time at least as recent as the time in the second message timestamp includes the steps of;

inspecting the logical clock associated with the node to determine a time indicated by said logical clock,

comparing the time indicated by said logical clock with a time indicated by the second message timestamp, and

if the time indicated by the second message timestamp is more recent than the time indicated by the logical clock, then setting the logical clock associated with the node to reflect a time that is at least as recent as the time reflected in the second message timestamp.

13. The system of claim 12, said plurality of nodes further comprising:

a central node, configured to transmit a first message timestamp representing time indicated by the logical clock associated with said central node to each other node of said plurality of nodes;

wherein each other node of said plurality of nodes is configured to transmit a first message timestamp representing time indicated by the logical clock associated with said node to said central node.

14. A computer readable medium having stored thereon sequences of instructions for maintaining a logical clock of a node in a multi-node parallel database system of a plurality of nodes, said sequences of instructions including instructions for performing the steps of:

maintaining a watermark that represents a time at least as recent as the latest time represented by the logical clocks on each of said plurality of nodes;

when the time represented on said logical clock is within a predetermined threshold of the time represented by said watermark, then increasing said watermark.

15. The computer readable medium of claim 14, wherein said sequence of instructions further includes instructions for performing the steps of:

stopping said logical clock before the step of causing said node to increase said watermark; and

resuming said logical clock, after the step of causing said node to increase said watermark.

16. The computer readable medium of claim 14, wherein said sequence of instructions further includes instructions for performing the steps of:

transmitting a first message that contains a first message timestamp representing time indicated by said logical clock;

receiving a second message from another node of said plurality of nodes, wherein said second message contains a second message timestamp; and

in response to receiving said second message, performing the steps of;

inspecting said logical clock to determine a time indicated by said logical clock,

comparing the time indicated by said logical clock with a time indicated by the second message timestamp, and

if the time indicated by the second message timestamp is more recent than the time indicated by said logical clock, then setting said logical clock to reflect a time that is at least as recent as the time reflected in the message timestamp.

17. The computer readable medium of claim 14, wherein said sequences of instructions further include instructions for performing the steps of:

detecting failure of one of said plurality of nodes; and

in response to detecting failure of said one of said plurality of nodes, causing a recovery node to perform the steps of;

setting the logical clock associated with the recovery node to reflect a time that is at least as recent as the time reflected in the watermark and

after setting the logical clock, causing said recovery node to recover the node that failed.

18. A computer readable medium having stored thereon sequences of instructions for maintaining a logical clock of a node in a distributed system of a plurality of nodes, wherein each node of said plurality of nodes is associated with a logical clock, said sequences of instructions including instructions for performing the step of:

causing said node to perform the following steps within each of a plurality of successive periods having a common prescribed duration;

(a) transmitting a first message that contains a first message timestamp representing time indicated by said logical clock to another node of said plurality of nodes;

(b) receiving a second message from another node of said plurality of nodes, wherein said second message contains a second message timestamp; and

(c) causing the logical clock associated with said node to reflect a time at least as recent as the time in the second message timestamp;

wherein the step of causing the logical clock associated with said node to reflect a time at least as recent as the time in the second message timestamp includes the steps of;

inspecting the logical clock associated with the node to determine a time indicated by said logical clock,

comparing the time indicated by said logical clock with a time indicated by the second message timestamp, and

if the time indicated by the second message timestamp is more recent than the time indicated by the logical clock, then setting the logical clock associated with the node to reflect a time that is at least as recent as the time reflected in the second message timestamp.

19. The computer readable medium of claim 18, wherein said sequence of instructions further includes instructions for performing the step of transmitting a first message timestamp representing time indicated by said logical clock to every other node of said plurality of nodes.

20. The computer readable medium of claim 18, wherein said sequence of instructions further includes instructions for performing the step of transmitting a first message timestamp representing time indicated by said logical clock to a central node.

21. The computer readable medium of claim 18, wherein said sequence of instructions further includes instructions for performing the steps of:

maintaining a watermark that represents a time at least as recent as the latest time represented by the logical clocks on each of said plurality of nodes;

when the time represented on said logical clock is within a predetermined threshold of the time represented by said watermark, then increasing said watermark;

detecting failure of one of said plurality of nodes; and

in response to detecting failure of said one of said plurality of nodes, causing a recovery node to perform the steps of;

setting the logical clock associated with the recovery node to reflect a time that is at least as recent as the time reflected in said watermark and

after setting the logical clock, causing said recovery node to recover the node that failed.

22. The computer readable medium of claim 21, wherein said sequence of instructions further includes instructions for performing the steps of:

stopping said logical clock before the step of increasing said watermark and resuming said logical clock after the step of increasing said watermark.