Data Pipelines

It's also knowing whether or not you can trust the data. You have to then if you fix a bug, you have to actually go back and repair data. And that's something that is really hard to solve without a really context to where a data-aware scheduler.