In addition to the below suggestion, If you’re developing simple ETL (i.e. no machine learning etc), I’d highly reccomend prototyping using DuckDB within knime (via DB nodes). You can then export the query and transpile it to another dialect ( e.g. databricks, snowflake , redshift and many others) pretty easily using SQLGlot python package. Then you can pretty much move the dev code to prod with little to no refactoring.