Sam works as a Network Analyst for an algorithmic trading firm. He obtained his Bachelors Degree in Information Technology from UMKC.
JPerf is a graphical front end for the popular network testing tool Iperf. Using JPerf you can quickly test a WAN or LAN connection to determine the maximum network throughput. The test results are automatically graphed and presented in a format that is easy to read. JPerf can also be used to detect packet loss, delay, jitter, and other common network problems.
JPerf provides many benefits over Iperf which is a command line only application. Besides being reliable and easy to use JPerf is completely free. The utility is fully open source and runs on both Windows or Linux systems.
Whether you're trying to diagnose a problem in your home network or troubleshoot the performance of a WAN link JPerf can handle the task.
Installing JPerf on Windows
JPerf requires that Java version 1.5 or newer is installed before it will run. You can visit Java.com to download the latest version or verify if it is correctly installed on your computer.
To get JPerf running you will need to download the jperf-2.0.2.zip file from the JPerf Google code page.
There isn't an installer so simply extract the contents of the zip file to a location on your computer such as C:\JPerf.
To launch the JPerf utility run jperf.bat.
Installing JPerf on Linux
On Linux the Java executable must be in your system path, to verify this type java -version in a shell. You should see output similar to what's below.
java version "1.6.0_18"
Java(TM) SE Runtime Environment (build 1.6.0_18-b07)
Java HotSpot(TM) 64-Bit Server VM (build 16.0-b13, mixed mode)
If you don't see this make sure Java is installed and is within your path.
Next download the JPerf zip file, the same archive is used for Windows or Linux. The folder can be extracted to any location, for example /usr/bin/jperf.
Within the jperf directory you downloaded is a script called jperf.sh. This script must be made executable by running chmod u+x jperf.sh from within the JPerf directory.
Finally, to launch the application run ./jperf.sh.
Setting up a JPerf Server
JPerf is designed to run as a client/server application. To run a test you will need to set up a JPerf server on your network. Then you can run a JPerf client from another location on the network which will connect to the remote server.
To start the JPerf server select the radio button labeled server then click Run IPerf. By default JPerf runs in TCP mode and listens on port 5001.
Connecting a client to the server
To connect to the JPerf server to run a test you'll need to first select the client radio button. In the server address field type in the IP address of the computer running the JPerf server. To begin the test click on run iPerf in the upper right hand corner of the app.
By default JPerf will run a 10 second TCP test using 1 stream. While the test is running the graph will update in real time to reflect the results.
There are several options that can be adjusted to modify the parameters of the test.
Application layer options
- Transmit - Run the test for a specified number of seconds, or until a certain amount of bytes have been transferred.
- Output Format - Test results can be changed to display bits, bytes, kbytes, etc.
- Report Interval - This adjusts how often the graph results are updated.
Transport layer options
There are several TCP options that can be modified such as buffer length, window size, and MSS. JPerf can also function in UDP mode, although the server must be operating in UDP mode in order for this test to work.
JPerf Tips and Tricks
Below are a few useful tips for improving your JPerf results.
- Use Parallel streams - The bandwidth of a single TCP session is limited by several factors. By using parallel streams you can easily saturate a very high bandwidth connection. In the JPerf client settings you can specify the number of streams to use. I've found 10 to be a good number.
- Run a Bi-Directional Test - By default JPerf transmits data from the client to the server. By selecting the dual testing mode under application layer options JPerf will send data in both directions at the same time.
- Use a representative file - JPerf has a cool ability that allows you to select a file to be transmitted to the server during the test. This function allows you to simulate a real world data transfer across your network in a controlled manner.
- Use JPerf to create iPerf commands - Since JPerf uses iPerf as a back end to run all of the tests you can use JPerf to help you build useful iPerf commands. Select the test options you want using the GUI and then copy the command it created from the box at the top of the application.
Expanding Your Knowledge
Now that you are familiar with JPerf I would recommend learning more about iPerf and how to run it from the command line. IPerf is designed for Linux but you can compile IPerf for Windows as well.
JPerf and Iperf tests can be run cross platform, eg: Linux to Windows or vice versa which makes it a highly useful tool for network testing. I've found both of these tools to be indispensable for diagnosing and troubleshooting network problems.
This article is accurate and true to the best of the author’s knowledge. Content is for informational or entertainment purposes only and does not substitute for personal counsel or professional advice in business, financial, legal, or technical matters.
sukanya on March 17, 2016:
Hello skear ,
I am using Jperf for testing my NTP server throughput. For sending NTP packet(48 bytes) i am using the " representative" file in .bin format. I just written 48 bytes in the file and i am able to send the packet through gui.
When i observe the packet in wireshark, after IP and UDP header its appending 38 bytes and then remaining data it is taking from the ".bin file".
From where this 38 bytes are including? Is i cannot send 48 bytes of data after UDP header?
Please suggest some solution for the above issue.
Thanks in advance.
Sam Kear (author) from Kansas City on May 28, 2013:
Hi Damian, Thanks for the feedback!
TCP sessions are often affected by what is called the long fat pipe problem and multiple streams are an easy way to overcome this issue.
There are two main factors that effect TCP throughput. The window size, and the latency of the path. To calculate the maximum possible throughput of a TCP session you can take the window size (in bits) and divide it by the latency of the network path.
For example, if the TCP window size is 32KB and the latency of the circuit is 5ms the maximum throughput of the session would be 51Mbps (32,000 * 8) / .005).
Jperf/iPerf also you to manually set a window size which can also help in testing. You can try using a large window size to improve the performance of a single session.
If you know the bandwidth of a line, and the latency you can calculate the optimal TCP window size using the following formula.
Bandwidth (in Bits per second) * Round trip latency (in seconds) = Window size in bits
Then multiply this calculated window size by 8 to convert it back to bytes.
So for example if the circuit is 10Gbps and the round trip latency is 1ms the optimal window size would be 1.25Mb. (10,000,000,000 *.001 = 10,000,000 bits / 8 = 1,2500,000 Kb)
Other factors come into play such as window scaling, slow start, lost packets, etc.
Hope this helps!
Damian on May 28, 2013:
Great write up, this helped me!
I have one question about the parallel streams, why do large bandwidth pipes require multiple streams to saturate the pipe?
sandeep on September 15, 2011: