Method and apparatus for providing failure detection and recovery with predetermined degree of replication for distributed applications in a network

Abstract

An application module (A) running on a host computer in a computer network is failure-protected with one or more backup copies that are operative on other host computers in the network. In order to effect fault protection, the application module registers itself with a ReplicaManager daemon process (,) by sending a registration message, which message, in addition to identifying the registering application module and the host computer on which it is running, includes the particular replication strategy (cold backup, warm backup, or hot backup) and the degree of replication associated with that application module. The backup copies are then maintained in a fail-over state according to the registered replication strategy. A WatchDog daemon (,), running on the same host computer as the registered application periodically monitors the registered application to detect failures. When a failure, such as a crash or hangup of the application module, is detected, the failure is reported to the ReplicaManager, which effects the requested fail-over actions. An additional backup copy is then made operative in accordance with the registered replication style and the registered degree of replication. A SuperWatchDog daemon process (,-,), running on the same host computer as the ReplicaManager, monitors each host computer in the computer network. When a host failure is detected, each application module running on that host computer is individually failure-protected in accordance with its registered replication style and degree of replication.