Include logic in either (or both) the orchestrator.lua and the EndpointProxy to favor processing smaller requests when running on a public cluster.
Periodically, the public cluster gets used by a group doing a large run - this run consumes all of the resources of the public cluster and prevents other people from having timely responses in their smaller requests.
Include logic in either (or both) the orchestrator.lua and the EndpointProxy to favor processing smaller requests when running on a public cluster.
Periodically, the public cluster gets used by a group doing a large run - this run consumes all of the resources of the public cluster and prevents other people from having timely responses in their smaller requests.