mmaybeno / py_profiling_talk

Simple talk about local profiling and simple solution for profiling web apps.

Quick profiling talk

July 2016 SD Python Meetup Talk
Based off blog post from nylas
Simple talk about local profiling and simple solution for profiling web apps.

cProfile

python app/cprofile_demo.py
         506 function calls in 0.000 seconds

   Ordered by: standard name

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        1    0.000    0.000    0.000    0.000 <string>:1(<module>)
        1    0.000    0.000    0.000    0.000 cprofile_demo.py:10(sort_fast)
        1    0.000    0.000    0.000    0.000 random.py:277(shuffle)
        1    0.000    0.000    0.000    0.000 {len}
        1    0.000    0.000    0.000    0.000 {method 'disable' of '_lsprof.Profiler' objects}
      499    0.000    0.000    0.000    0.000 {method 'random' of '_random.Random' objects}
        1    0.000    0.000    0.000    0.000 {range}
        1    0.000    0.000    0.000    0.000 {sorted}


         500006 function calls in 0.568 seconds

   Ordered by: standard name

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        1    0.008    0.008    0.568    0.568 <string>:1(<module>)
        1    0.002    0.002    0.560    0.560 cprofile_demo.py:16(sort_slow)
        1    0.262    0.262    0.294    0.294 random.py:277(shuffle)
        1    0.000    0.000    0.000    0.000 {len}
        1    0.000    0.000    0.000    0.000 {method 'disable' of '_lsprof.Profiler' objects}
   499999    0.032    0.000    0.032    0.000 {method 'random' of '_random.Random' objects}
        1    0.010    0.010    0.010    0.010 {range}
        1    0.255    0.255    0.255    0.255 {sorted}

Line and memory profiler

pip install -r requirements.txt
./line_mem_demo.sh 
+ kernprof -l -v app/fast_demo.py
Wrote profile results to fast_demo.py.lprof
Timer unit: 1e-06 s

Total time: 0.000627 s
File: app/fast_demo.py
Function: sort_fast at line 6

Line #      Hits         Time  Per Hit   % Time  Line Contents
==============================================================
     6                                           @profile
     7                                           def sort_fast():
     8         1           13     13.0      2.1      number_list = range(500)
     9         1          524    524.0     83.6      random.shuffle(number_list)
    10         1           90     90.0     14.4      sorted(number_list)

+ python -m memory_profiler app/fast_demo.py
Filename: app/fast_demo.py

Line #    Mem usage    Increment   Line Contents
================================================
     6   10.461 MiB    0.000 MiB   @profile
     7                             def sort_fast():
     8   10.473 MiB    0.012 MiB       number_list = range(500)
     9   10.473 MiB    0.000 MiB       random.shuffle(number_list)
    10   10.473 MiB    0.000 MiB       sorted(number_list)


+ kernprof -l -v app/slow_demo.py
Wrote profile results to slow_demo.py.lprof
Timer unit: 1e-06 s

Total time: 0.910497 s
File: app/slow_demo.py
Function: sort_slow at line 6

Line #      Hits         Time  Per Hit   % Time  Line Contents
==============================================================
     6                                           @profile
     7                                           def sort_slow():
     8         1         9739   9739.0      1.1      number_list = range(500000)
     9         1       593153 593153.0     65.1      random.shuffle(number_list)
    10         1       307605 307605.0     33.8      sorted(number_list)

+ python -m memory_profiler app/slow_demo.py
Filename: app/slow_demo.py

Line #    Mem usage    Increment   Line Contents
================================================
     6   10.707 MiB    0.000 MiB   @profile
     7                             def sort_slow():
     8   26.230 MiB   15.523 MiB       number_list = range(500000)
     9   26.230 MiB    0.000 MiB       random.shuffle(number_list)
    10   33.832 MiB    7.602 MiB       sorted(number_list)

Build docker images for web profiling using docker-compose

./build_images.sh

Boot up flask app and nylas collection/visualizer

docker-compose up

Run Wrk benchmarking tool

./make_wrk.sh

# https://github.com/wg/wrk
$ wrk -t1 -c1 -d5s http://localhost:8000/fast
Running 5s test @ http://localhost:8000/fast
  1 threads and 1 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    18.56ms    1.75ms  27.19ms   60.82%
    Req/Sec    53.04      5.74    60.00     60.00%
  268 requests in 5.05s, 45.02KB read
Requests/sec:     53.07
Transfer/sec:      8.91KB
$ wrk -t1 -c1 -d5s http://localhost:8000/slow
Running 5s test @ http://localhost:8000/slow
  1 threads and 1 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    31.66ms    1.23ms  37.21ms   79.87%
    Req/Sec    30.96      3.91    40.00     82.00%
  159 requests in 5.09s, 26.71KB read
Requests/sec:     31.23
Transfer/sec:      5.25KB

View flask app endpoints

http://localhost:8000/fast
http://localhost:8000/slow

View flask app stack samples

http://localhost:8000

View flamegraph of stack samples

http://localhost:8001