Re: Help? Where?
The easiest support is to join the iperf-users mailing list, and post your
questions.
As for your results and subsequent checks with Chariot:
I am not sure why you are having this sensitivity to packet size I have
not seen this before, though you could search the archives at
http://archive.ncsa.uiuc.edu/lists/iperf-users/ . As for the Chariot
results I am pretty sure that they use their own TCP/IP stack and by-pass
the WinXP one. Correct me if I am wrong but Chariot is mostly marketed at
testing IP Stack compliance and operation, which implies to me that they
are sending through their own stack. If this is the case then the problem
could be the windows stack. Iperf uses the windows stack just as any other
application might. Therefore if there is a problem with their stack it
could cause this. Also I find that a UDP test is not the best test for a
customer to perform since almost NO network applications send bulk UDP. A
true test for what one could expect from a connection is a TCP test which
translates to WWW and FTP speeds that can be expected. Granted this will
always be less than the purchased "link speed" however it is the "goodput"
that can be expected as a customer. I would love to see ISPs start selling
using this sort of number, but alas it will never catch on. Sorry I could
not be more of a help.
Kevin
On Wed, 7 Jul 2004, Mitch Kutzko wrote:
> >From: "Richard Clark" <rclark --at-- airgain.com>
> >To: <mitch --at-- dast.nlanr.net>
> >Subject: Help? Where?
> >Date: Wed, 7 Jul 2004 12:47:13 -0700
> >
> >
> >Mitch,
> >
> >Sorry to bother you directly, but I cant seem to find a support contact
> anywhere&
> >Tried the support/contact page, but the comment line only takes, well, one
> line.
> >Tried the email link, but that doesnt work either (dast --at-- nlanr.net)...
> >Hoping your address works! J
> >
> >Also tried to find anything about KNOWN PROBLEMS , but search doesnt work,
> >and there is zilch in the iperf documentation&
> >
> >Heres the original text/problem&
> >
> >Hi!
> >
> >A customer of ours reported some throughput problems, which we
> subsequently investigated, and found to be something odd in the iperf
> testing tool. We
> >are hoping you might have some answers/guidance for us.
> >
> >Namely, we are running testing between two WinXP machines, via the
> >dos command prompt, using iperf u s i 5 and iperf u c
> ><ipaddr> -i 5 t 30 b 10m , through a pair of wireless Ethernet bridges.
> >The results that we get using different packet sizes -l x are very
> strange.
> >For packet 1100 bytes or smaller, we get throughput about 6.2 mbits/sec.
> >
> >For packets 1200 to 1490 bytes, we get throughput of about 1.7mbits/sec.
> >
> >Yes, even the default UDP packet size of 1470 gets throughput of about
> 1.5mbits/sec.
> >
> >And for packets 1500 bytes or more, we again get throughput of 6.2 mbits/sec.
> >
> >We tried to reproduce this behavior with the Chariot testing tool, but
> varying the
> >UDP datagram sizes there did not have the same result. In other words,
> Chariot gave
> >about 6.2mbits throughput for all the above packet sizes&
> >
> >So, my question is, are there known problem/quirks with 1200-1490 byte
> packets?
> >
> >Have you seen this problem before? Any clues/help you can give us?
> >
> >Thanks!!!
> >
> >Richard H. Clark
> >Airgain, Inc.
> >
> >C:\IPERF>iperf -s -u -i 5 -l 1460
> >------------------------------------------------------------
> >Server listening on UDP port 5001
> >Receiving 1460 byte datagrams
> >UDP buffer size: 8.00 KByte (default)
> >------------------------------------------------------------
> >[1968] local 192.168.0.21 port 5001 connected with 192.168.0.190 port 1552
> >[ ID] Interval Transfer Bandwidth Jitter Lost/Total
> Datagrams
> >[1968] 0.0- 5.0 sec 1.03 MBytes 1.72 Mbits/sec 8.992 ms 1302/ 2039 (64%)
> >[1968] 5.0-10.0 sec 1.05 MBytes 1.76 Mbits/sec 9.008 ms 1400/ 2155 (65%)
> >[1968] 10.0-15.0 sec 1.04 MBytes 1.74 Mbits/sec 5.347 ms 1387/ 2134 (65%)
> >[1968] 15.0-20.0 sec 1.02 MBytes 1.71 Mbits/sec 7.762 ms 1393/ 2127 (65%)
> >[1968] 20.0-25.0 sec 1.05 MBytes 1.76 Mbits/sec 5.505 ms 1402/ 2154 (65%)
> >[1968] 25.0-30.0 sec 1.02 MBytes 1.72 Mbits/sec 11.355 ms 1384/ 2120 (65%)
> >[1968] 0.0-30.3 sec 6.26 MBytes 1.74 Mbits/sec 12.375 ms 8347/12844 (65%)
> >
> --
> Mitch Kutzko | mitch --at-- dast.nlanr.net | mitch --at-- ncsa.uiuc.edu | 217-333-1199
> Project: http://dast.nlanr.net/ | Personal: http://hobbes.ncsa.uiuc.edu/
>