On Composition of Checkpoint and Recovery Protocols for Distributed Systems

Rollback recovery has been studied as a low-cost fault tolerance mechanism for ensuring dependability of critical distributed applications. There is a rich variety of recovery protocols proposed in literature and they are broadly classified as checkpoint-based recovery protocols and message-log based recovery protocols. In this paper we attempt to model composition of protocol and check whether such composition is consistent with recovery. The composition of protocols in important in a system whether resources are hierarchically organized, for example grid or cloud systems.