Architecture and Nomenclature¶

Diagram showing the relation between server and client processes.

Server / Main Process¶

The server is running in the cluster_utils main process which is started by grid_search or hp_optimization on the login node of the cluster. This process submits jobs to the cluster’s scheduler, monitors them and collects the results.

While the jobs register with the server and send their results, the server itself doesn’t actually send any data to the jobs directly. It only monitors them based on the data they sent and on information obtained from the cluster system.

Client / Job Script¶

The jobs are executed by the cluster system on the compute nodes of the cluster. They execute the user script that is specified with the script_relative_path setting. This script is expected to call cluster_utils.initialize_job() in the beginning, to register at the server (so the server knows that the job started), and cluster_utils.finalize_job() at the end to send its results to the server.