Benchmark: Socketify vs FastWSGI #110
Description
Machine: AMD EPYC 7543 @ 3.7GHz, Debian 11, Python 3.9
Payload for testing: https://github.com/MiloszKrajewski/SilesiaCorpus/blob/master/xml.zip (651 KiB)
Server test app: https://gist.github.com/remittor/c9411e62b5ea4776200bee288a331016
FastWSGI project: https://github.com/jamesroberts/fastwsgi
Socketify (multi-threaded)
> python3 server.py -g si -t8
> wrk -t1 -c1 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
1 threads and 1 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 49.60us 61.19us 4.78ms 99.30%
Req/Sec 20.79k 2.32k 26.18k 78.15%
312288 requests in 15.10s, 46.76MB read
Requests/sec: 20681.11
Transfer/sec: 3.10MB
> wrk -t8 -c8 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
8 threads and 8 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 61.00us 64.76us 5.04ms 98.97%
Req/Sec 16.78k 3.24k 20.77k 70.45%
2016690 requests in 15.10s, 301.95MB read
Requests/sec: 133554.59
Transfer/sec: 20.00MB
> wrk -t8 -c128 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
8 threads and 128 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 454.64us 109.56us 3.29ms 72.81%
Req/Sec 35.11k 1.94k 42.48k 67.61%
4216557 requests in 15.10s, 631.33MB read
Requests/sec: 279246.40
Transfer/sec: 41.81MB
> python3 server.py -g si -t8 -f xml.zip -b
> wrk -t1 -c1 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
1 threads and 1 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 230.31us 458.91us 15.63ms 99.34%
Req/Sec 4.95k 1.12k 6.28k 56.29%
74349 requests in 15.10s, 46.20GB read
Requests/sec: 4923.96
Transfer/sec: 3.06GB
> wrk -t8 -c8 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
8 threads and 8 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 534.77us 445.86us 11.07ms 93.31%
Req/Sec 1.98k 764.95 3.56k 61.92%
238315 requests in 15.10s, 148.09GB read
Requests/sec: 15782.65
Transfer/sec: 9.81GB
Socketify (single-threaded)
> python3 server.py -g si -t1
> wrk -t1 -c1 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
1 threads and 1 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 54.19us 58.65us 4.83ms 99.43%
Req/Sec 18.85k 763.04 19.83k 84.00%
281228 requests in 15.00s, 42.11MB read
Requests/sec: 18748.14
Transfer/sec: 2.81MB
> wrk -t8 -c8 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
8 threads and 8 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 281.61us 73.64us 4.76ms 92.65%
Req/Sec 3.57k 320.89 6.26k 82.82%
427728 requests in 15.10s, 64.04MB read
Requests/sec: 28326.63
Transfer/sec: 4.24MB
> wrk -t8 -c128 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
8 threads and 128 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 3.98ms 239.06us 8.56ms 87.62%
Req/Sec 4.04k 114.86 4.36k 59.17%
482031 requests in 15.01s, 72.17MB read
Requests/sec: 32119.73
Transfer/sec: 4.81MB
> python3 server.py -g si -t1 -f xml.zip -b
> wrk -t1 -c1 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
1 threads and 1 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 342.26us 1.07ms 23.18ms 98.60%
Req/Sec 4.26k 1.53k 6.32k 50.99%
63969 requests in 15.10s, 39.75GB read
Requests/sec: 4236.57
Transfer/sec: 2.63GB
> wrk -t8 -c8 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
8 threads and 8 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 2.11ms 452.35us 7.02ms 89.65%
Req/Sec 474.80 68.24 626.00 74.58%
56754 requests in 15.01s, 35.27GB read
Requests/sec: 3781.08
Transfer/sec: 2.35GB
FastWSGI (single-threaded)
> python3 server.py -g fw -t1
> wrk -t1 -c1 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
1 threads and 1 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 53.38us 208.03us 10.83ms 99.68%
Req/Sec 21.72k 0.88k 22.89k 89.40%
326314 requests in 15.10s, 31.74MB read
Requests/sec: 21610.83
Transfer/sec: 2.10MB
> wrk -t8 -c8 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
8 threads and 8 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 617.78us 5.29ms 94.86ms 98.96%
Req/Sec 7.39k 1.41k 9.34k 75.91%
888591 requests in 15.10s, 86.44MB read
Requests/sec: 58847.30
Transfer/sec: 5.72MB
> wrk -t8 -c128 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
8 threads and 128 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 2.08ms 3.44ms 66.57ms 98.49%
Req/Sec 9.19k 1.48k 38.75k 94.01%
1098912 requests in 15.10s, 106.90MB read
Requests/sec: 72777.04
Transfer/sec: 7.08MB
> python3 server.py -g fw -t1 -f xml.zip -b
> wrk -t1 -c1 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
1 threads and 1 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 180.08us 226.19us 9.09ms 99.12%
Req/Sec 5.94k 1.23k 7.08k 81.46%
89346 requests in 15.10s, 55.52GB read
Requests/sec: 5917.17
Transfer/sec: 3.68GB
> wrk -t8 -c8 -d15 http://127.0.0.1:5000
Running 15s test @ http://127.0.0.1:5000
8 threads and 8 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 1.50ms 793.92us 46.30ms 86.38%
Req/Sec 676.81 155.80 0.93k 75.75%
80895 requests in 15.02s, 50.27GB read
Requests/sec: 5387.46
Transfer/sec: 3.35GB
Conclusion:
Project socketify is very fast when using multithreading. But in single-threaded mode, server FastWSGI turned out to be faster.