How mlcp Adjusts Client Concurrency
When running an import job, mlcp periodically send polling requests to the server through the XCC layer to obtain the maximum server threads. When the DHS cluster adds more dynamic e-nodes, server has more available concurrency. Then mlcp decides whether to scale-out or scale-in, its own thread pool based on the result.
The following command line options can be used to tune this process:
-max_thread_percentage: The percentage (between 0 and 100) of maximum available server threads mlcp will use to run import jobs.-polling_period: The time interval (in minutes) mlcp sends the polling requests to the server.-polling_init_delay: The initial delay (in minutes) before mlcp starts sending the polling requests.