Method and apparatus for fault-tolerant, scalable and non-blocking three-phase flushing for committing database transactions in a cluster of multiprocessors

Abstract

An apparatus, method and medium for implementing fault-tolerant three-phase flushing for database transaction commits in a cluster including multiple nodes of shared-memory processors. The apparatus, method and medium include executing the primary of a commit coordinator process pair on a first of the nodes and executing its backup on a second node, executing the primary of a resource manager process pair on a third node and executing its backup on a fourth, and executing the primary of a Log process pair on a fifth node and executing its backup on a sixth. The first and second nodes as distinct from each other, as are the third and fourth nodes, as are the fifth and sixth nodes.

Claims aren’t available for this document.