Team, I have drafted NIP-26 [1] proposing the introduction of a new (Experimental) Scheduling Strategy for Processors. In this new Scheduling Strategy, it would be up to the framework to automatically determine how many threads a given Processor gets, and what the Run Duration should be. The framework would have the freedom to update the values dynamically in order to ensure adequate resources without overwhelming the system. Often we see this become a difficult part for users to tune, and appropriate values sometimes vary from environment to environment (going from a 2 CPU-core container to a 24 CPU-core container means the values set for one environment are not optimal for the other).
I won’t be surprised if there’s a decent bit of discussion here so I’ll be happy to engage in the DISCUSS thread or via the Jira. If all sounds positive I’ll be happy to call a vote after discussions settle. Thanks -Mark [1] https://issues.apache.org/jira/browse/NIP-26
