Cluster-wide read-copy update system and method

• US 8,126,843 B2
• Filed: 11/30/2004
• Issued: 02/28/2012
• Est. Priority Date: 11/30/2004
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: McKenney, Paul E., Satran, Julian
• Primary Examiner(s): Alam, Hosain
• Assistant Examiner(s): JOHNSON, JOHNESE T
• Application Number:
  US10/999,833
• Publication Number:
  US 20060117072A1
• Time in Patent Office: 2,646 Days
• Field of Search: 707/2, 707/201, 707/610, 707/705
• US Class Current: 707/610
• CPC Class Codes: G06F 16/273 Asynchronous replication or reconciliation

Abstract

A system, method and computer program product for synchronizing updates to shared mutable data in a clustered data processing system. A data element update operation is performed at each node of the cluster while preserving a pre-update view of the shared mutable data, or an associated operational mode, on behalf of readers that may be utilizing the pre-update view. A request is made for detection of a grace period, and grace period detection processing is performed for detecting when the cluster-wide grace period has occurred. When it does, a deferred action associated with the update operation it taken, such as removal of a pre-update view of the data element or termination of an associated mode of operation.

Claims

1. A method for synchronizing data updates to shared mutable data maintained at each node in a clustered data processing system, comprising:

performing a data element update operation on a data element maintained at each node of said clustered data processing system by updating said data element at each node while preserving a pre-update view of said shared mutable data, or perserving an associated operational mode, on behalf of readers that are utilizing said pre-update view or said operational mode, wherein said readers may continue operations without regard to said update operation;

requesting detection of a grace period at each node;

detecting when a cluster-wide grace period has occurred that guarantees a grace period has occurred at each node and said readers will no longer be utilizing said pre-update view or said operational mode; and

performing a deferred action associated with said update operation that removes said pre-date view or changes said operational mode following detection of said cluster-wide grace period.

2. The method in accordance with claim 1, wherein said update operation is initially performed at an update-initiating node followed by other nodes in said clustered data processing system.

3. The method in accordance with claim 2, wherein said update-initiating node advises said other nodes of said update.

4. The method in accordance with claim 1, wherein said requesting comprises registering a callback at each node.

5. The method in accordance with claim 1, wherein said cluster-wide grace period detection is performed by an update-initiating node.

6. The method in accordance with claim 5, wherein said cluster-wide grace period detection is performed by said update-initiating node receiving reports from other nodes in said clustered data processing system passing through local grace periods or quiescent states.

7. The method in accordance with claim 1, wherein said cluster-wide grace period detection is performed at all nodes of said clustered data processing system.

8. The method in accordance with claim 7, wherein said cluster-wide grace period detection is performed by each node of said clustered data processing system maintaining a count of nodes passing through local grace periods or quiescent states.

9. The method in accordance with claim 1, wherein said cluster-wide grace period is based on all nodes in said clustered data processing system passing through a local grace period.

10. The method in accordance with claim 1, wherein said cluster-wide grace period is based on all nodes in said clustered data processing system passing through a local quiescent state.