>I very much doubt that fping is the right tool for this task.

As far as I understand this, floodping (on linux) sends out as much
icmp-echo requests as possible (number of packets send depend on the
source system).
If the number/s multiplied with the size of each packet is greater than
the physical capability of the line, you should be able to flood it.
If this is the only traffic on the line, the iftop far side will display
the received bandwidths (thru echo requests).

On a 100Mb/s - Lan-Connection, I measure about 96Mb/s with iftop far
side thru that method.

Why do you think flooding the line with icmp packets is an invalid test?


>> Before you start - yes, I tried iperf, but I couldn't saturate the
link
>> with iperf - not even with udp. Not sure, why this didn't work, but
ping
>> -f from a linux system did the trick though.
>
>I'd suggest using that or netperf.  And if you're not able to saturate
>the link with those conventional testing tools, then (quite simply)
your
>OpenSolaris system is not able to saturate the link.  The problem is
>very unlikely to be a lack of "fping."

This might drift a bit off-topic. But I hope this discussion is in the
interest of everybody.

First - I tried iperf from a linux system with udp - to become
independent of return traffic (ACKs) with the following parameters on a
6Mb/s-Wan connection :

ipfer -u -b 10M

this generated about 3.6 Mb/s Traffic on the link.

I then stopped iperf and used floodping on the same WAN connection with
the same partners and created a load of 5.5 Mb/s (got the Bandwidth -
readings from snmp from the receiving Router - so this was the overall
Bandwidth of the link).

So to me it looks like iperf wasn't capable of flooding the link,
although the 10M is well above the 6M available. Did I use the wrong
parameters to get the result, I tried to create?

BTW - I sometimes even got negative results (<0) with ipfer - this sure
looks like a bug to me and basically destroyed my trust in iperf. 

Cheers,
Kai


_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to