parsl.executors.high_throughput.manager_selector.RandomManagerSelector

class parsl.executors.high_throughput.manager_selector.RandomManagerSelector[source]

Returns a shuffled list of interesting_managers

By default this strategy is used by the interchange. Works well in distributing workloads equally across all availble compute resources. The random workload strategy is not effective in conjunction with elastic scaling behavior as the even task distribution does not allow the scaling down of blocks, leading to wasted resource consumption.

__init__()[source]

Methods

__init__()

sort_managers(ready_managers, manager_list)

Sort a given list of managers.

sort_managers(ready_managers: Dict[bytes, ManagerRecord], manager_list: Set[bytes]) List[bytes][source]

Sort a given list of managers.

Any operations pertaining to the sorting and rearrangement of the interesting_managers Set should be performed here.