Thanks for an interesting discussion.

>From: cisco-nsp [] on behalf of Saku Ytti 
>On 12 August 2017 at 18:21, Raymond Burkholder <> wrote:
>> I have successfully run iperf bidirectionally in tcp as well as udp and hit
>> link limits, even on smaller, lower capacity linux based boxes.
>On what packet sizes? What link speeds? Linux UDP socket performance
>is terrible, even with rescvmmsg sendmmsg which iperf does not
>utilise, the performance is bad. XEON grade server CPU won't congest
>1GE single dir - 1.48Mpps, without loss.
>If you don't care/look at loss, or use low pps, it's different.
>If you use TCP you're measuring the host stacks TCP implementation,
>and you have no visibility on network quality, because packet loss is
>hidden from you.

There is now a version 3 of iperf available.  Most noticeable improvement from 
this perspective is that in TCP mode it shows the current window size and the 
number of re-transmissions at each reporting interval.  This makes it much 
easier to understand the behaviour of the transfer... You see the 
re-transmissions caused by any packet loss in real time, and its immediate 
impact on the TCP throughput.  I find this very helpful.

The packet loss is no longer hidden like it was with iperf_v2.

iperf3 doesn't interoperate with the old iperf (v2) - you need a version 3 at 
both ends.


Tim Rayner
Optical Engineer, AARNet Pty Ltd
street address: Building 9, Banks Street, YARRALUMLA  ACT  2600
postal address:  GPO Box 1559, CANBERRA   ACT   2601
t. +61 2 6222 3575     f. +61 2 6222 3535     e.
Network Faults (24x7): 1300 APL NOC (1300 275 662) or

This email and any files transmitted with it are confidential, and the rights 
of confidentiality in such information are not waived or lost by its mistaken 
delivery to you.  Any dissemination, copying, use or disclosure of the email 
and/or such files without the permission of AARNet, or the sender, is strictly 
prohibited.  If you have received this email in error, please contact the 
sender immediately and delete all copies of this transmission.
cisco-nsp mailing list
archive at

Reply via email to