Browse Prior Art Database

RECOVERY MECHANISM IN LONG RUNNING CONVERSATIONAL APPLICATION

IP.com Disclosure Number: IPCOM000014417D
Original Publication Date: 2000-Jul-01
Included in the Prior Art Database: 2003-Jun-19

Publishing Venue

IBM

Abstract

In a transaction processing environment, the underlying resource managers, e.g., Databases, file systems, maintain a log of inter- actions of an application in order to achieve atomicity of a set of interactions of the application with the resource manager(s); Upon a failure the resource managers use undo and/or redo oper- ations based on the log to bring the resources to a consistent state. In addition, a recoverable application and/or an under- lying monitor checkpoints the states of the recoverable applica- tion at all commit points for recovery upon a failure. Disclosed is a method via which the state of a long running application that interacts with autonomous external entities (e.g., autonomous applications, resource managers, etc. where interactions can not be undone) can be recreated by an underlying monitor without checkpointing various intermediate states of the application. The monitor maintains a log of interactions with all other autonomous entities on behalf of the application, and upon a failure, replays the external events (e.g., request/reply from autonomous entities) in order to drive the application from an initial state to its before failure state. DETAILED DESCRIPTION OF THE FRAMEWORK: Coyote architecture