Build Spark data pipelines for large datasets.
Recent revisions are kept. Restoring one creates a new revision.
