Testing Wi-Fi with OSS Tools
Download as ODP, PDF0 likes441 views
This document summarizes tools and techniques for open source network testing, including testing routers and Wi-Fi networks with multiple concurrent clients to evaluate performance under real-world conditions. It describes using tools like net-hydra, netburn, and whenits to automate testing across multiple client devices and collect throughput and latency statistics. The document advocates an approach of testing networks with multiple concurrent activities like downloading, browsing, VoIP calls, and streaming to evaluate how equipment handles collision domains under more challenging real-world loads.
![enter the hydra
net-hydra -c {configfile} [-t {testname}]
net-hydra requires a config file specifying [clients], [directives], and optionally
[aliases], which it uses to schedule jobs to run simultaneously in the very near future
on multiple client machines.
# sample nethydra.conf
[clients]
local0 = root@127.0.0.1
local1 = root@127.0.0.2
[directives]
# These are the command(s) to be run on each client device. Directives
# beginning with $ reference lines from the [aliases] section. Any use
# of $when, either here or in [aliases], will be replaced with the
# scheduled execution time, in Unix epoch seconds; $testname is replaced
# with the --testname argument passed to net-hydra on the command-line.
local0 = $4kstream
local1 = /usr/bin/touch /tmp/test-$testname-$when.txt
[aliases]
# Defining aliases here keeps [directives] cleaner, so you can
# see what you're doing a little more easily.
#
$4kstream = netburn -u http://127.0.0.1/1M.bin -r 25 -o /tmp/$testname-$when.csv](https://image.slidesharecdn.com/2017-ato-network-testing-171127155738/85/Testing-Wi-Fi-with-OSS-Tools-15-320.jpg)
![ab wouldn’t do… so, netburn
netburn -u {url}
[-r {rate limit} ] ... specified in Mbps (default none)
[-t {seconds} ] ... time to run test, default 30
[-o {filespec} ] ... output filespec for CSV report
[--no-header ] ... suppress header row of CSV output
[-h {hostname} ] ... override system-defined hostname
in CSV output
[-c {concurrency} ] ... number of concurrent URL fetches to
make per 'page' fetch
[-q ] ... quiet (suppress all but CSV output)
[-ifinfo {iface} ] ... output detailed wifi info about {iface}
[--usage ] ... you're looking at this right now](https://image.slidesharecdn.com/2017-ato-network-testing-171127155738/85/Testing-Wi-Fi-with-OSS-Tools-16-320.jpg)
![/usr/bin/at only granular to 1m... wtf?
whenits [-d] {time} {/path/to/command {arg1 arg2 arg3 ...} }
-d ... optional: daemonize (fork and run in
background). You'll need this if you're
scheduling commands that need to run after
the current shell closes.
time ... Required. Execution time may be specified
in absolute epoch seconds, or relative to
current time - eg now+5s, now+10m, now+2h.
command ... the command to be scheduled at {time},
along with any arguments to be passed to it.
whenits {time}, run {command}](https://image.slidesharecdn.com/2017-ato-network-testing-171127155738/85/Testing-Wi-Fi-with-OSS-Tools-18-320.jpg)
Testing Wi-Fi with OSS Tools
- 1. OSS Network Testing breaking the network, not the bank This presentation is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License. (C) 2017 [email protected] Jim Salter Technomancer, Mercenary Sysadmin, Small Business Owner Today's slides can be found at: http://openoid.net/presentations/
- 2. A couple of years ago I got really interested in routers.
- 3. I didn’t trust consumer stuff … so I built my own.
- 4. Once I built it, I had to test it. (e-peen ain’t gonna measure itself)
- 5. iPerf, big numbers, go home? NOPE - real protocols needed
- 6. I tried using CIFS tests... mount -t cifs ; time cat /stuff
- 7. but ppl relate more to HTTP ApacheBench to the rescue!
- 8. ab, smaller numbers, go home! me@banshee:~$ ab -c10 -t3 http://192.168.0.30/1M.bin This is ApacheBench, Version 2.3 <$Revision: 1706008 $> Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/ Licensed to The Apache Software Foundation, http://www.apache.org/ Benchmarking 192.168.0.30 (be patient) Completed 5000 requests Completed 10000 requests Finished 11486 requests (better, but limited ability to model)
- 9. gigabit consumer switch gigabit Ubuntu router
- 10. gigabit consumer switch “gigabit” consumer router
- 11. “… but you couldn’t pay me enough to care about Wi-Fi!” well ACTUALLY...
- 12. Breakthrough tri-band Wi-Fi technology with amazing combined wireless speeds of up to 5,332 Mbps!!!1one
- 13. we need to move past the single client model
- 14. I always want free speech But this time, I also needed the beer! ixChariot == $$$
- 15. enter the hydra net-hydra -c {configfile} [-t {testname}] net-hydra requires a config file specifying [clients], [directives], and optionally [aliases], which it uses to schedule jobs to run simultaneously in the very near future on multiple client machines. # sample nethydra.conf [clients] local0 = [email protected] local1 = [email protected] [directives] # These are the command(s) to be run on each client device. Directives # beginning with $ reference lines from the [aliases] section. Any use # of $when, either here or in [aliases], will be replaced with the # scheduled execution time, in Unix epoch seconds; $testname is replaced # with the --testname argument passed to net-hydra on the command-line. local0 = $4kstream local1 = /usr/bin/touch /tmp/test-$testname-$when.txt [aliases] # Defining aliases here keeps [directives] cleaner, so you can # see what you're doing a little more easily. # $4kstream = netburn -u http://127.0.0.1/1M.bin -r 25 -o /tmp/$testname-$when.csv
- 16. ab wouldn’t do… so, netburn netburn -u {url} [-r {rate limit} ] ... specified in Mbps (default none) [-t {seconds} ] ... time to run test, default 30 [-o {filespec} ] ... output filespec for CSV report [--no-header ] ... suppress header row of CSV output [-h {hostname} ] ... override system-defined hostname in CSV output [-c {concurrency} ] ... number of concurrent URL fetches to make per 'page' fetch [-q ] ... quiet (suppress all but CSV output) [-ifinfo {iface} ] ... output detailed wifi info about {iface} [--usage ] ... you're looking at this right now
- 17. feeling the burn you@box:~$ netburn -u http://server/128K.bin -r 10 Fetching http://server/128K.bin at 10 Mbps for 30 seconds. Throughput so far: 10.01 Mbps Seconds remaining: 0 Time elapsed: 30.2 seconds Number of pages fetched: 301 Total data fetched: 37.6 MB Mean page length fetched: 1048576 KB Page length maximum deviation: 0 KB Throughput achieved: 10 Mbps Mean latency: 41.65 ms Worst latency: 190 ms 99th percentile latency: 51 ms 95th percentile latency: 46 ms 90th percentile latency: 45 ms 75th percentile latency: 43 ms Median latency: 41 ms Min latency: 35 ms
- 18. /usr/bin/at only granular to 1m... wtf? whenits [-d] {time} {/path/to/command {arg1 arg2 arg3 ...} } -d ... optional: daemonize (fork and run in background). You'll need this if you're scheduling commands that need to run after the current shell closes. time ... Required. Execution time may be specified in absolute epoch seconds, or relative to current time - eg now+5s, now+10m, now+2h. command ... the command to be scheduled at {time}, along with any arguments to be passed to it. whenits {time}, run {command}
- 19. OK, let’s pull some NUMBERS! (and tell a story with them)
- 20. single-client testing ho, hum… boring
- 21. single-client testing ho, hum at long range
- 22. now let’s test them with our multiclient test model Downloading Browsing VoiP...ing Streaming
- 23. wi-fi: it’s all about the collision domains aggressive spectrum use auto band steering – – vs – – manual band allocation radios and access points
- 24. multi-client testing hard mode (bandsteer) Archer C7 Archer C5400 Linksys EA8300 Netgear R8000
- 25. simplifying the charts: web browsing fails first image credit: US State Department – Doug Thompson
- 26. multi-client testing easy mode (manual)
- 27. multi-client testing hard mode (bandsteer)
- 28. bring on the mesh! (but remember how flaky R8000 was in bedroom2.)
- 29. Boo Plume. Yay Ubiquiti! Collision domains: they matter.
- 30. GTFO with your Wi-Fi P-t-P > broadcast. Always.