Towards Unifying Fault and Change ~a~a~~~~~~~
Jeff Kramer, Jeff Magee, Andrew Young · 1990
A basic model for dynamic change management has been proposed which permits changes to be specified declaratively at the configuration level, in terms of structure only. Component nodes can be created and deleted and interconnections made and broken. Rules are provided which permit configuration management to identify the affected part of the system and to generate the required operational changes from this specification. These operations change the state of the selected component nodes to that suitable for reconfiguration, perform the particular structural changes, and provide the sequencing of the actions. A node responds by performing node level actions to maintain local and intemode consistency. This model also provides a sound basis for the management of faults where failure of a component node is modelled as arbitrary deletion. Incomplete transactions of the connected nodes must be aborted. Recovery actions can be used to restore local and internode consistency. Reconfiguration changes can then be specified as part of the change transaction niggered by a f;iilure. This paper describes the first steps towards unifying the management of both dynamic change and fault recovery.