Glossary · Automation software engineering and architecture
Checkpoint (software)
Also known as: Checkpoint, Checkpointing
German: Prüfpunkt
In software engineering, a checkpoint is a saved snapshot of a program's or system's state at a defined point, from which execution can be resumed or restored after a failure or interruption, instead of starting again from the beginning.
- Software engineering
In one sentence
A checkpoint is a saved snapshot of a program's state from which execution can resume after a failure or interruption.
Example
A batch controller writes a checkpoint after each completed phase, so after a power failure the operator can resume the recipe at the last confirmed phase instead of restarting it.
How it applies
- Engineering: Checkpoints store the minimum state needed to resume safely: step numbers, counters, setpoints and pending actions. They must be written atomically, so a failure during writing does not leave a corrupt checkpoint.
- Operation: Resuming from a checkpoint is only safe if the physical process is still in the state the checkpoint describes. After a long stop, material, temperatures or positions may have changed, so resume logic often needs operator confirmation and plausibility checks.
- Documentation: Operating manuals should explain what happens after an interruption: which data is kept, how to resume and when a restart from the beginning is required. Service documentation should describe where checkpoints are stored and how to clear them.
Checkpoint vs. backup
A checkpoint captures the running state of a process for fast recovery. A backup preserves data and configuration for long-term restoration. Both are needed, and neither replaces the other.
In other contexts, 'checkpoint' can also mean a review milestone in a project or a trained model snapshot in machine learning.