reg : multicast throughput

[Sandeep Subramaniam <sandeeps --at-- ittc.ku.edu>]

Kevin,

Thanks for your information.

I used

iperf -c 225.0.0.5 -u -l 10 -b 110m -t 1 -T 5 // at the sender

and

iperf -s -u -B 225.0.0.5 // at the receiver


and measured throughput at the receiver for different message sizes.

As u had memtioned , to make the sender send as fast as possible, I have a
the send bandwidth set to an unrealistic high value of 110 Mbps on a 100
Mpbs Ethernet.

Thanks a lot !

Sandeep.

On Fri, 20 Dec 2002, Kevin Gibbs wrote:

> Yeah you are missing the -b command line arguement that determines the
> bandwidth to use. This is a UDP so there are no algorithms to "probe" for
> available bandwidth and send at that or any congestion control mechanisms
> to back off if we run out of bandwidth. I am not really sure what you are
> trying to do, but you could put an unrealistic value for -b like 101M when
> the link is only 100Mbit/s this will effectly send as fast as possible
> from the sender and the network will probably drop some packets along the
> way. If you are trying to see what the linux box can handle as far as CPU
> utilization then you can increment -b and watch the utilization on the
> linux box.

Sandeep Subramaniam
Graduate Research Assistant
ITTC, Raymond Nichols Hall,
University of Kansas