Deadline scheduling for Linux

I see your point. I don't see how a execution time would be completely irrelevant. You could potentially use that information for ordering, such as priorities small task above big tasks in case large task doesn't complete on time or priorities large tasks for long uninterrupted processing with small tasks filling in the spaces.

Execution time would be difficult at best to predict but could be used as a guestimate and record actually min/max execution times.