UCL
 

Personal Miscellaneous TCP/IP GRID Quality of Service Multi-Cast

 

lan wan

 

Iperf Tests Page

iperf if a tool to send real tcp traffic across a network. it is used to determine actively, what the goodput/throughput of a network is. It gives a good indication of the transfer rate available to eg ftp.

Description of iperf. here. High level description of what iperf does.
Duration of test against throughput. here. This experiment determines that there is an upper bound on the calculation of the amount of data transferred. This 4Gb limit forces iperf to only transmit data in the remainder of 4Gb if doing a sized transfer, and causes the reporting of the achieved throughput to be only calculated from the last remainder of the 4Gb.
Affect of increasing socket buffer size. here. Experiment to show the affect of increasing the socket buffer size of an iperf sender while keeping the reciever at a large socket buffer size. Results show that a double plateau is discovered, most probably as a result of the 2.4 linux kernel's implementation of basic autotuning. Further tests show that both the sender and receiver windows reported by iperf are incorrect (version 1.6.1) and that the iperf server has a physical limit of a 2meg window size.
Affect of cpu load depending on buffer size. here.

Experiments to try to quantify the load effect of running iperf transfers. This experiment needs further work.

Affect of reciever buffers and solution. here. This experiment was initially run to determine whether the socket buffer sizes of the reciever were correctly set on an old version of iperf and linux.
Variation in throughput here. Quantifies the variation in iperf throughput. In WAN tests, the variation is thought to be due to cross traffic. This experiment tries to replicate the variation on back to back tests.
Comparison of iperf results and web100 results. here. Comparision of the amount of data sent as reported by web100 and with iperf. Special considertion of the 4gb limit is explored.
   

A list of iperf problems is found here.

 

 

Wed, 23 July, 2003 13:07 Previous PageNext Page

 
 
    email me!
2001-2003, Yee-Ting Li, email: ytl@hep.ucl.ac.uk, Tel: +44 (0) 20 7679 1376, Fax: +44 (0) 20 7679 7145
Room D14, High Energy Particle Physics, Dept. of Physics & Astronomy, UCL, Gower St, London, WC1E 6BT