And now for the final conversion we have to perform: jumping from supervisors down to slots. As before with components and their executors, we need this because the cluster assigns executors at the slot level, not the supervisor level.
There are a few things to do at this point; we have broken the process down into smaller methods to preserve readability. The main steps we need to perform are as follows:
Find out which slots we can assign to, given a list of supervisors for a type. This is simply the case of using a for loop that collects all supervisors' slots, and then returning as many of the slots as are requested by the topology.
Divide the executors awaiting scheduling for the type into even groups across ...
