Skip to content

Latest commit

 

History

History

benchmarks

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 
 
 

apachebench-graph

Helper scripts for apachebench to automate the plotting of test results.

Requirements

  • apachebench
  • bash
  • gnuplot
  • sed

Usage

There are three helper script included:

  • ab_graph.sh: Run an apache bench test and create plot result graphs using gnuplot.
  • ab_graph_merge.sh: Merge various ab_graph.sh results graphs into a single one.
  • ab_graph_scale.sh: Helper script for ab-graph that can run multiple configurable ab_graph.sh runs in one go

ab_graph.sh

This script has the following options:

Usage: $0 OPTIONS

OPTIONS:
-c    Concurrent connections  (default: 1)
-k    Enable keepalive        (default: no)
-E    Extra parameters
-n    Number of requests      (default: 1)
-u    Url to test             (mandatory)
-h    Print help.
-V    Debug mode.

This script will plot apachebench results using gnuplot, and store them in $PWD/results/$WESITEURL/$DATE/. The script will create the plot files for gnuplot using the templates in the templates directory. It will also save apachebench output in a file called summary.txt.

For example, this command:

./ab-graph.sh -u http://www.testsite.com/ -n 500 -c 100 -k

Will create the following results:

Test results

Test summary results

  • summary.txt
This is ApacheBench, Version 2.3 <$Revision: 1843412 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking www.testsite.com (be patient)


Server Software:        nginx/1.10.3
Server Hostname:        www.testsite.com
Server Port:            80

Document Path:          //
Document Length:        4 bytes

Concurrency Level:      100
Time taken for tests:   26.012 seconds
Complete requests:      500
Failed requests:        0
Non-2xx responses:      500
Keep-Alive requests:    0
Total transferred:      200000 bytes
HTML transferred:       2000 bytes
Requests per second:    19.22 [#/sec] (mean)
Time per request:       5202.324 [ms] (mean)
Time per request:       52.023 [ms] (mean, across all concurrent requests)
Transfer rate:          7.51 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    1   1.3      0       7
Processing:   185 4926 1699.9   4718   10431
Waiting:      180 4925 1699.8   4718   10431
Total:        185 4926 1700.2   4718   10434

Percentage of the requests served within a certain time (ms)
  50%   4718
  66%   5203
  75%   5587
  80%   5781
  90%   7518
  95%   8573
  98%   9825
  99%  10228
 100%  10434 (longest request)

Plot templates

You can use templates to customize the graphs. Please refer to gnuplot documentation for that. These are the default templates:

Plotted results

Results

Results graph

Percentiles

Percentages graph

ab_graph_merge.sh

With this script you can merge various results from ab_graph.sh into one single graph, useful to compare results. For example, this command will merge two test results into one:

./ab-graph_merge.sh $PWD/results/www.testsite.com/2020-04-23-00-30-12 $PWD/results/www.testsite.com/2020-04-22-23-34-50

The resulting merged results graph looks like this:

Results

Merged results graph

Percentiles

Merged percentiles graph

ab_graph_scale.sh

This helper script can do multiple runs of ab_graph.sh in one go. It has many configurable parameters:

Usage: ./ab-graph_scale.sh OPTIONS

OPTIONS:
-c    Start concurrent connections at        (default: 5)
-e    Stop concurrent connections at         (default: 25)
-k    Enable keepalive connections           (default: no)
-s    Concurrent connections increment step  (default: 5)
-n    Number of requests                     (default: 500)
-u    Url to test                            (mandatory)
-w    Wait time between tests in seconds.    (default: 60s)

For example:

./ab-graph_scale.sh -c 10 -e 50 -k -s 10 -n 1000 -u https://www.test.com -w 30

Will do 1000 requests starting with 10 concurrent connections up to 50 in increments of 10, using keep alives and waiting 30 seconds between tests.