Skip to content

Instantly share code, notes, and snippets.

@jhochwald

jhochwald/iperf3_mini_workshop.md

Last active January 31, 2019 03:42
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save jhochwald/810bd54ab78d96a151b17489c70226bc to your computer and use it in GitHub Desktop.
Save jhochwald/810bd54ab78d96a151b17489c70226bc to your computer and use it in GitHub Desktop.
Download ZIP
iperf3 a TCP, UDP, and SCTP network bandwidth measurement tool (From a mini workshop)
Raw
iperf3_mini_workshop.md

iperf3 a TCP, UDP, and SCTP network bandwidth measurement tool

iperf is a free tool for active measurements of the maximum achievable bandwidth on IP networks. It supports tuning of various parameters related to timing, protocols, and buffers.

For each test it reports the measured throughput / bitrate, loss, and other parameters.

Command line arguments

The following arguments are the one that we used in the workshop. All will work in iperf3 version 3.1, or newer. All of them will work on Windows, Linux, and macOS. There are a few others, that might be plattform specific.

Client related arguments

The -c / --client argument let iPerf in client mode, connecting to an iPerf server running on host.

The -p / --port argument specifies server port for the server to listen on and the client to connect to, default is 5201.

The -t / --time argument specifies the test duration time in seconds, default is 10 secs.

The -i / --interval argument indicates the interval in seconds between periodic bandwidth reports, default is zero.

The -u / --udp argument uses UDP rather than TCP. (TCP is the default)

The -r / --reverse argument run everything in reverse mode (server sends, client receives).

The --udp-counters-64bit argument supports very long-running UDP tests, which could cause a counter to overflow

The --cport argument specifies the client-side port. Useful for tracing or Firewall tests.

The -k / --blockcount argument defines the number of blocks (packets) to transmit.

The -l / --length argument define the length of buffers to read or write, default is 128 KB for TCP, 8 KB for UDP.

The -P / --parallel argument defines the number of simultaneous connections to make to the server, default is 1.

The -V / --verbose argument give more detailed output

The -J / --json argument creates output in JSON format

The -4 / --version4 argument only use IPv4.

The -6 / --version6 argument only use IPv6.

The -w / --window argument sets the socket buffer sizes to the specified value. For TCP, this sets the TCP window size.

The -M / --set-mss argument attempts to set the TCP maximum segment size (MSS). The MSS is usually the MTU - 40 bytes for the TCP/IP header. For ethernet, the MSS is 1460 bytes (1500 byte MTU).

The -O / --omit argument omits the first n seconds of the test, to skip past the TCP slowstart period.

The -Z / --zerocopy argument uses a "zero copy" method of sending data, such as sendfile(2), instead of the usual write(2). This uses much less CPU.

The -N / --no-delay argument set the TCP no delay option, disabling Nagle's algorithm. Normally this is only disabled for interactive applications like telnet.

The -f / --format argument specifies the format to print bandwidth numbers in. Supported formats are:

  • k = Kbits/sec
  • K = KBytes/sec
  • m = Mbits/sec
  • M = MBytes/sec

Sample:

iperf3 -c 10.26.22.110 -p 5201 -t 20 -i 2 -f m

Sample:

iperf3 -c 10.26.22.110 -p 5201 -t 20 -i 2 -f m -R

Sample:

iperf3 -c 10.26.22.110 -p 5201 -t 120 -i 4 -f M

Sample:

iperf3 -c 10.26.22.110 -p 5201 -t 120 -i 4 -f M -R

Sample:

iperf3 -c bouygues.iperf.fr -p 5200 -t 20 -i 2 -f m

Sample:

iperf3 -c bouygues.iperf.fr -p 5200 -t 20 -i 2 -f m -R

Sample:

iperf3 -c 10.26.22.110 -p 5201 -t 120 -i 4 -R -f M -u

Sample

See how much bandwidth can be pushed using UDP

iperf3 -c 10.26.22.110 -p 5201 -u -t 30 -b 50m
iperf3 -c 10.26.22.110 -p 5201 -u -t 30 -b 50m -R
iperf3 -c 10.26.22.110 -p 5201 -u -t 30 -b 100m
iperf3 -c 10.26.22.110 -p 5201 -u -t 30 -b 100m -R
iperf3 -c 10.26.22.110 -p 5201 -u -t 30 -b 300m
iperf3 -c 10.26.22.110 -p 5201 -u -t 30 -b 300m -R
iperf3 -c 10.26.22.110 -p 5201 -u -t 30 -b 500m
iperf3 -c 10.26.22.110 -p 5201 -u -t 30 -b 500m -R
iperf3 -c 10.26.22.110 -p 5201 -u -t 30 -b 600m
iperf3 -c 10.26.22.110 -p 5201 -u -t 30 -b 600m -R

Sample

See how much bandwidth can be pushed using TCP

iperf3 -c 10.26.22.110 -p 5201 -t 30 -b 50m
iperf3 -c 10.26.22.110 -p 5201 -t 30 -b 50m -R
iperf3 -c 10.26.22.110 -p 5201 -t 30 -b 100m
iperf3 -c 10.26.22.110 -p 5201 -t 30 -b 100m -R
iperf3 -c 10.26.22.110 -p 5201 -t 30 -b 300m
iperf3 -c 10.26.22.110 -p 5201 -t 30 -b 300m -R
iperf3 -c 10.26.22.110 -p 5201 -t 30 -b 500m
iperf3 -c 10.26.22.110 -p 5201 -t 30 -b 500m -R
iperf3 -c 10.26.22.110 -p 5201 -t 30 -b 600m
iperf3 -c 10.26.22.110 -p 5201 -t 30 -b 600m -R

Sample

Parallel tests, single client, TCP window moving up.

If you start seeing a lot of Retrs/drops you reached the limit. I use this kind of tests for WiFi or WAN connections to find the limits of them.

If your Client CPU usage get's to hig, try the -Z argument. If the local CPU is overloaded, this might givr you the wrong results.

2 concurrent threads
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 2 -w 10k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 2 -w 10k -R
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 2 -w 100k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 2 -w 100k -R
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 2 -w 300k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 2 -w 300k -R
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 2 -w 400k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 2 -w 400k -R
5 concurrent threads
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 5 -w 10k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 5 -w 10k -R
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 5 -w 100k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 5 -w 100k -R
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 5 -w 300k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 5 -w 300k -R
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 5 -w 400k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 5 -w 400k -R
10 concurrent threads
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 10 -w 10k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 10 -w 10k -R
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 10 -w 100k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 10 -w 100k -R
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 10 -w 300k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 10 -w 300k -R
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 10 -w 400k
iperf3 -c 10.26.22.110 -p 5201 -t30 -P 10 -w 400k -R

Sample

See the impact when disabling no delay. Guess?

iperf3 -c 10.26.22.110 -p 5201 -t30 -N
iperf3 -c 10.26.22.110 -p 5201 -t30 -N -R
iperf3 -c 10.26.22.110 -p 5201 -t30
iperf3 -c 10.26.22.110 -p 5201 -t30 -R

Server related arguments

The -s / --server argument starts iPerf in server mode. (This will only allow one iperf connection at a time)

The -p / --port argument specifies server port for the server to listen on and the client to connect to, default is 5201.

Start iPerf Server on Port 5201

iperf3 -s -p 5201

Start iPerf Server on Port 5200

iperf3 -s -p 5200

Start iPerf Server on Port 5200

Same as above, but as daemon (Centos Linux)

sudo -u nobody iperf3 -s -p 5200 -D >/dev/null 2>&1

Project home

You should visit the iPerf Hoempage!

Download

You will find a lot of pre-compiled binaries in the download section of the Project.

Documentation

There is an excelent documentation.

GitHub

There is a GitHub page for this project.

Bug Tracker

iPerf3 issue tracker on GitHub: https://github.com/esnet/iperf/issues

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment