What are the main parameters we usually set for DAG tasks, and why should the graph be structured this way (top-level DAG logic)?