[c-nsp] Testing bandwidth

[Will Hargrave]

Ed Butler wrote:
> I would say FTP is going to be pretty unreliable as a test method on
> high throughput links, because of the TCP receive window. It will be
> fine for testing E1s etc though :)
> 
> See:
> http://www.babinszki.com/Networking/Max-Ethernet-and-TCP-Throughput.html

That document looks pretty old and I don't think really reflects 
enhancements like RFC1323 window scaling, SACK etc, all of which ship in 
modern OSs.

Lots of e-science bulk transfer testing is done using ftp or iperf...

Not sure what problems the OP has with iperf, whether he's finding it 
buggy or simply isn't producing the wanted results ;-)

http://www.caida.org/tools/taxonomy/performance.xml has a useful list of 
tools.

Will