Hi! In short: for UDP testing, use server side (that is: the side that receives traffic) reports.
Why: UDP is unacknowledged mode, so the sending party doesn't know if the packed got delivered or not. Therefore sender can not produce detailed report. And the same goes for server side report at the end of test: it can happen that client doesn't receive that information. This could be due to many reasons: - server never receives last few test packets so it doesn't know when to send such report - server sends report but client doesn't receive it due to congestion (and subsequent packet drop) in the link between - ... If you want to have detailed statistics, you should use TCP testing, which has its own drawbacks (eg. inability to show maximum throughput when speed in other direction is much lower, slow starts after retransmissions, retransmissions themself, ...) or UDP testing where statistics are actually gathered on the receiving side. We (mobile broadband operator) do the later. BR, Metod Sergey Naumov je dne 29/08/11 14:30 napisal-a: > Hello. > > I use iperf 2.0.4 for udp testing (2.0.5 is broken for udp- 1) very > big pkt loss is reported in contrast with 2.0.4, 2) test hangs in the > middle). > > I use -i1 option to be able to count minimum-maximum values of > traffic, but report of traffic sent to server is not corrected by > datagram-loss percent, so I can't use this values directly. I have to > wait until test finish and then look at server report where I can see > datagram loss percent of client->server flow. Then I can correct > values of sent traffic of client->server flow. It leads to incorrect > minimum-maximum values for this flow, because I have only average > datagram loss. It is acceptable, but there is one more problem: > sometimes server reports are not received and iperf hangs. I will > reluctant to dig inside iperf code to make it work but maybe somebody > can say where this bug may reside to make my life easier. > > BTW, it is very inconvinient to get statictics from server not about > last test, but about the whole time server runs. I can get proper > datagram loss statistics if I restart the server (inconvinient), but > information of Mbit/s in this case is irrelevant. > > Thanks in advance, > Sergey Naumov. > > ------------------------------------------------------------------------------ > EMC VNX: the world's simplest storage, starting under $10K > The only unified storage solution that offers unified management > Up to 160% more powerful than alternatives and 25% more efficient. > Guaranteed. http://p.sf.net/sfu/emc-vnx-dev2dev > _______________________________________________ > Iperf-users mailing list > Iperf-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/iperf-users > ------------------------------------------------------------------------------ EMC VNX: the world's simplest storage, starting under $10K The only unified storage solution that offers unified management Up to 160% more powerful than alternatives and 25% more efficient. Guaranteed. http://p.sf.net/sfu/emc-vnx-dev2dev _______________________________________________ Iperf-users mailing list Iperf-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/iperf-users
