libzmq/perf/generate_graphs.py

50 lines
1.7 KiB
Python
Raw Normal View History

#!/usr/bin/python
#
# This script assumes that a CSV file produced by "generate_csv.sh" is provided as input
#
# configurable values:
INPUT_FILE_PUSHPULL_TCP_THROUGHPUT="pushpull_tcp_thr_results.csv"
INPUT_FILE_PUSHPULL_INPROC_THROUGHPUT="pushpull_inproc_thr_results.csv"
INPUT_FILE_PUBSUBPROXY_INPROC_THROUGHPUT="pubsubproxy_inproc_thr_results.csv"
INPUT_FILE_REQREP_TCP_LATENCY="reqrep_tcp_lat_results.csv"
# dependencies
import matplotlib.pyplot as plt
import numpy as np
# functions
def plot_throughput(csv_filename, title):
message_size_bytes, message_count, pps, mbps = np.loadtxt(csv_filename, delimiter=',', unpack=True)
plt.semilogx(message_size_bytes, pps / 1e6, label='PPS [Mmsg/s]', marker='x')
plt.semilogx(message_size_bytes, mbps / 1e3, label='Throughput [Mb/s]', marker='o')
plt.xlabel('Message size [B]')
plt.title(title)
plt.legend()
plt.show()
plt.savefig(csv_filename.replace('.csv', '.png'))
def plot_latency(csv_filename, title):
message_size_bytes, message_count, lat = np.loadtxt(csv_filename, delimiter=',', unpack=True)
plt.semilogx(message_size_bytes, lat, label='Latency [us]', marker='o')
plt.xlabel('Message size [B]')
plt.title(title)
plt.legend()
plt.show()
plt.savefig(csv_filename.replace('.csv', '.png'))
# main
plot_throughput(INPUT_FILE_PUSHPULL_TCP_THROUGHPUT, 'ZeroMQ PUSH/PULL socket throughput, TCP transport')
plot_throughput(INPUT_FILE_PUSHPULL_INPROC_THROUGHPUT, 'ZeroMQ PUSH/PULL socket throughput, INPROC transport')
plot_throughput(INPUT_FILE_PUBSUBPROXY_INPROC_THROUGHPUT, 'ZeroMQ PUB/SUB PROXY socket throughput, INPROC transport')
plot_latency(INPUT_FILE_REQREP_TCP_LATENCY, 'ZeroMQ REQ/REP socket latency, TCP transport')