Optimize end to end launch time from OSS to MLOps

[chaoyanghe]

[chaoyanghe]

monitor_service(queue): req_json = queue.get(block=False)

def start_monitor(queue): writer = multiprocessing.Process(target=monitor_service, args=(queue,)) writer.start()

job_queue = multiprocessing.Queue() job_queue.put(req_json)

start_monitor(job_queue)

[chaoyanghe]

Server agent start server: 09:49:47 ~ 09:49:53, about 6 seconds，

[chaoyanghe]

Server: download_mnist -> load_synthetic_data: 09:51:28 ~ 09:51:43, about 15 seconds (MNIST.zip downloaded) load_synthetic_data -> load_partition_data_mnist: 09:51:43 ~ 09:52:09, about 26 seconds create_model -> connection_ready: 09:52:09 ~ 09:52:10, about 1 seconds

Client: load_synthetic_data -> load_partition_data_mnist: 9:54:05 ~ 9:54:59, about 54 seconds load_partition_data_mnist -> connection_ready: 09:55:00 ~ 09:55:01, about 1 second