Checkpointing is a technique for inserting fault tolerance into computing systems. It basically consists on storing a snapshot of the current application state, and uses it for restarting the execution in case of failure. It is saving the program state, usually to stable storage, so that it may be reconstructed later in time. Checkpointing provides the backbone for rollback recovery (fault-tolerance), playback debugging, process migration, and job swapping. It mainly focuses on fault-tolerance, process migration, and the performance of checkpointing on all computational platforms from uniprocessors to supercomputers.

Checkpointing and restart has been one of the most widely used techniques for fault tolerance in large parallel applications. By periodically saving application status to permanent storage (disk or tape), the execution can be restarted from the last checkpoint if system faults occur. It is an effective approach to tolerating both hardware and software faults. For example, a user who is writing a long program at a terminal can save the input buffer occasionally to minimize the rewriting caused by failures that affect the buffer.