iperf is a 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 bandwidth, loss, and other parameters.

This version, sometimes referred to as iperf3, is a redesign of an original version developed at NLANR/DAST. iperf3 is a new implementation from scratch, with the goal of a smaller, simpler code base, and a library version of the functionality that can be used in other programs. iperf3 also a number of features found in other tools such as nuttcp and netperf, but were missing from the original iperf. These include, for example, a zero-copy mode and optional JSON output. Note that iperf3 is NOT backwards compatible with the original iperf.

Primary development for iperf3 takes place on CentOS Linux, FreeBSD, and MacOS X. At this time, these are the only officially supported platforms, however there have been some reports of success with OpenBSD, Android, other Linux distributions.

iperf3 is principally developed by ESnet / Lawrence Berkleley National Laboratory. It is released under a three-clause BSD license.

For more information see: https://github.com/esnet/iperf

Downloads of iperf3 are available at:

http://stats.es.net/software/

To check out the most recent code, do:

git clone https://github.com/esnet/iperf.git

• libuuid: this is not installed by default for Debian/Ubuntu Systems to install: apt-get install uuid-dev

./configure; make; make install

(Note: If configure fails, try running ./bootstrap.sh first)

iperf3 includes a manual page listing all of the command-line options. The manual page is the most up-to-date reference to the various flags and parameters.

For sample command line usage, see:

http://fasterdata.es.net/performance-testing/network-troubleshooting-tools/iperf-and-iperf3/

Using the default options, iperf is meant to show typical well designed application performance. 'Typical well designed application' means avoiding artificial enhancements that work only for testing (such as splice()'ing the data to /dev/null). iperf does also have flags for 'extreme best case' optimizations but they must be explicitly activated.

These flags include:

-Z, --zerocopy            use a 'zero copy' sendfile() method of sending data
-A, --affinity n/n,m      set CPU affinity

Before submitting a bug report, try checking out the latest version of the code, and confirm that its not already fixed. Then submit to: https://github.com/esnet/iperf/issues

New options:

-V, --verbose             more detailed output than before
-J, --json                output in JSON format
-Z, --zerocopy            use a 'zero copy' sendfile() method of sending data
-O, --omit N              omit the first n seconds (to ignore slowstart)
-T, --title str           prefix every output line with this string
-F, --file name           xmit/recv the specified file
-A, --affinity n/n,m      set CPU affinity (Linux and FreeBSD only)
-k, --blockcount #[KMG]   number of blocks (packets) to transmit (instead 
                          of -t or -n)
-L, --flowlabel           set IPv6 flow label (Linux only)

Changed flags:

-C, --linux-congestion    set congestion control algorithm (Linux only)
                          (-Z in iperf2)

Deprecated options:

Not planning to support these iperf2 flags. If you really miss these options, please submit a request in the issue tracker:

-d, --dualtest           Do a bidirectional test simultaneously
-r, --tradeoff           Do a bidirectional test individually
-T, --ttl                time-to-live, for multicast (default 1)
-x, --reportexclude [CDMSV]   exclude C(connection) D(data) M(multicast) 
                              S(settings) V(server) reports
-y, --reportstyle C      report as a Comma-Separated Values

Also deprecated is the ability to set the options via environment variables.

• UDP performance: iperf2/iperf3 both only are only about 50% as fast as nuttcp in UDP mode. We are looking into this, but in the meantime, if you want to get UDP above 5Gbps, we recommend using nuttcp instead (http://www.nuttcp.net/).

• Interval reports on high-loss networks: The way iperf3 is currently implemented, the sender write command will block until the entire block has been written. This means that it might take several seconds to send a full block if the network has high loss, and the interval reports will have widely varying interval times. We are trying to determine the best solution to this, but in the meantime, try using a smaller block size if you get strange interval reports. For example, try "-l 4K".

• The -Z flag sometimes hangs on OSX.

• On OpenBSD, the server seems to require a "-4" argument, implying that it can only be used with IPv4.

• When specifying the TCP buffer size using the "-w" flag on Linux, Linux doubles the value you pass in. (You can see this using iperf3's debug flag). But then the CWND does not actually ramp up to the doubled value, but only to about 75% of the doubled value. This appears to be by design.