Generation data sets have predictable backup and availability requirements in comparison to other batch data sets. We provide guidelines for developing management classes for both sets of batch data.