The shared-memory timeslice interface supports multiple clients in a variety of configurations. However, it does not currently support the distribution of timeslices to multiple clients that are equivalent. This is the case in a scenario with a number of timeslice worker processes running in parallel on a processing node.
An idea to support this use case includes:
Define worker groups.
Each worker group is identified by an additional tag, which is communicated on connection.
The members of the worker group share all other timeslice selection parameters and the corresponding queue.
The worker group mostly acts like a single worker.
Each timeslice shall be given to only one member of the worker group, based on the order of requests.
The shared-memory timeslice interface supports multiple clients in a variety of configurations. However, it does not currently support the distribution of timeslices to multiple clients that are equivalent. This is the case in a scenario with a number of timeslice worker processes running in parallel on a processing node.
An idea to support this use case includes: