From: Paul H. <Pau...@no...> - 2005-10-13 21:59:59
|
Depending on the hardware speed, operating system, interface, and buffer tuning, it is quite common to have larger UDP buffer sizes attain higher throughput than buffer size matched to the interface MTU. (Reduced CPU load is what I usually find, it's pretty easy to see at 1Gbit/sec and above.) Najaf, you may also find the data resources and tools at the Internet2 End2End Performance Initiative useful. [bwctl and the I2 monitoring results in particular] http://e2epi.internet2.edu/ http://e2epi.internet2.edu/pipes/ami/pmp-info.html Paul Hyder NOAA Earth System Research Laboratory, Global Systems Division Boulder, CO yang hua wrote: > I have similar questions. I tried iperf (udp) in a > peer-to-peer 802.11g network, when I use a much longer > udp packet size (32k) I can get higher throughput than > using shorter udp packet size (1400bytes). I am kind > of confused. If the IP MTU is 1500 bytes, 1400 bytes > of udp packet size should be optimum to achieve the > highest MAC throughput. But the lab test shows the > contrary. > > Anyway has any idea? > > Thanks > > > --- "Najaf A. Shah" <naj...@gm...> wrote: > > >>Hi, >> has anyone here tested iperf 'out in the field' >>using different >>packet sizes and other parameters? If so, I'd >>really like to hear >>about your experience and look at your data if >>possible. >> >>I'm working on an university research project and my >>work is not for profit... >> >>Thanks >> >>Najaf |