I have a program that looks like this... What would be the best way to process this in dask, so it works on both single machine and an auto-scaling cluster (say GCloud
