1. Check out our resources on running on Ray for distributed computing: https://www.getdaft.io/projects/docs/en/latest/user_guide/integrations/ray.html it should be a simple 1-line function call to connect to a cluster! 2. You can either run distributed for more resources, or you can also try running on a local Ray cluster which will allow you to run out of core. Check out some more tips on processing large datasets: https://www.getdaft.io/projects/docs/en/latest/user_guide/poweruser/memory.html 3. No! We use arrow formats under the hood, but we don’t currently leverage any parts of Datafusion