This project is no longer actively maintained. While existing releases remain available, there are no planned updates, bug fixes, new features, or security patches. Users should be aware that vulnerabilities may not be addressed.

TorchServe Benchmark on gpu

Date: 2022-03-22 04:27:24

TorchServe Version: master

eager_mode_mnist

version	Benchmark	Batch size	Batch delay	Workers	Model	Concurrency	Requests	TS failed requests	TS throughput	TS latency P50	TS latency P90	TS latency P99	TS latency mean	TS error rate	Model_p50	Model_p90	Model_p99	predict_mean	handler_time_mean	waiting_time_mean	worker_thread_mean	cpu_percentage_mean	memory_percentage_mean	gpu_percentage_mean	gpu_memory_percentage_mean	gpu_memory_used_mean
master	AB	1	100	4	.mar	10	10000	0	2345.99	4	5	6	4.263	0.0	1.04	1.15	1.53	1.06	1.02	1.93	0.28	0.0	0.0	0.0	0.0	0.0
master	AB	2	100	4	.mar	10	10000	0	3261.31	3	4	5	3.066	0.0	1.36	1.91	2.18	1.45	1.41	0.17	0.44	0.0	0.0	0.0	0.0	0.0
master	AB	4	100	4	.mar	10	10000	0	2457.64	4	6	7	4.069	0.0	1.89	2.2	2.96	1.97	1.94	0.53	0.59	0.0	0.0	0.0	0.0	0.0
master	AB	8	100	4	.mar	10	10000	0	1640.2	5	9	11	6.097	0.0	2.95	3.15	3.43	3.0	2.96	1.06	0.8	0.0	0.0	0.0	0.0	0.0
master	AB	1	100	8	.mar	10	10000	0	3444.57	3	3	4	2.903	0.0	1.32	1.68	1.87	1.37	1.34	0.08	0.46	0.0	0.0	0.0	0.0	0.0
master	AB	2	100	8	.mar	10	10000	0	3275.88	3	4	5	3.053	0.0	1.61	2.23	2.51	1.72	1.68	0.01	0.55	0.0	0.0	0.0	0.0	0.0
master	AB	4	100	8	.mar	10	10000	0	2346.15	4	6	8	4.262	0.0	2.01	2.42	3.19	2.1	2.06	0.57	0.57	0.0	0.0	0.0	0.0	0.0
master	AB	8	100	8	.mar	10	10000	0	1572.82	5	9	12	6.358	0.0	3.09	3.39	4.7	3.15	3.11	1.1	0.82	0.0	0.0	0.0	0.0	0.0

eager_mode_vgg16

version	Benchmark	Batch size	Batch delay	Workers	Model	Concurrency	Requests	TS failed requests	TS throughput	TS latency P50	TS latency P90	TS latency P99	TS latency mean	TS error rate	Model_p50	Model_p90	Model_p99	predict_mean	handler_time_mean	waiting_time_mean	worker_thread_mean	cpu_percentage_mean	memory_percentage_mean	gpu_percentage_mean	gpu_memory_percentage_mean	gpu_memory_used_mean
master	AB	1	100	4	.mar	100	10000	0	277.64	353	384	478	360.178	0.0	13.27	14.49	18.55	13.61	13.57	343.11	0.35	69.2	11.3	22.25	12.4	2004.0
master	AB	2	100	4	.mar	100	10000	0	284.7	344	377	462	351.248	0.0	25.69	29.79	49.7	26.86	26.82	320.57	0.84	33.3	11.29	16.25	12.39	2002.0
master	AB	4	100	4	.mar	100	10000	0	298.66	331	355	386	334.831	0.0	50.61	54.65	72.63	51.69	51.64	278.95	1.33	66.7	11.63	16.0	12.81	2070.0
master	AB	8	100	4	.mar	100	10000	0	302.97	321	367	401	330.066	0.0	100.17	108.43	134.97	102.03	101.97	222.5	2.62	0.0	12.1	15.25	13.4	2166.0

scripted_mode_vgg16

version	Benchmark	Batch size	Batch delay	Workers	Model	Concurrency	Requests	TS failed requests	TS throughput	TS latency P50	TS latency P90	TS latency P99	TS latency mean	TS error rate	Model_p50	Model_p90	Model_p99	predict_mean	handler_time_mean	waiting_time_mean	worker_thread_mean	cpu_percentage_mean	memory_percentage_mean	gpu_percentage_mean	gpu_memory_percentage_mean	gpu_memory_used_mean
master	AB	1	100	4	.mar	100	10000	0	282.06	351	368	430	354.53	0.0	13.18	13.91	18.68	13.41	13.37	337.73	0.33	80.0	11.32	23.25	12.4	2004.0
master	AB	2	100	4	.mar	100	10000	0	288.03	345	363	406	347.18	0.0	25.68	29.08	40.61	26.53	26.49	316.93	0.83	37.5	11.31	16.5	12.39	2002.0
master	AB	4	100	4	.mar	100	10000	0	296.25	332	356	447	337.552	0.0	50.72	55.09	84.0	52.09	52.04	281.21	1.34	0.0	11.63	16.0	12.81	2070.0
master	AB	8	100	4	.mar	100	10000	0	301.07	324	367	407	332.147	0.0	100.49	109.71	136.18	102.69	102.63	223.7	2.59	0.0	0.0	0.0	0.0	0.0