3 issues




I have three issue with iperf, one of which is new to 1.6.3 and the others
appear in both 1.6.3 and earlier versions 

On windows we always use the pre-compiled version. For Solaris I have used
the pre-compiled 1.1.1 version and both the pre-compiled and locally
compiled 1.6.3 version.

These tests are between a Win32 machine and a Solaris8 machine 

First the new one 
When starting the Win32 machine with the -w 64K option I get a warning and
the window size does not change. This worked fine under 1.2

D:\installers\iperf-1.6.3-win32>iperf -s -p 10000 -i 1 -w 64k 
------------------------------------------------------------ 
Server listening on TCP port 10000 
TCP window size:  8.0 KByte (WARNING: requested 64.0 KByte) 
------------------------------------------------------------ 

D:\installers\iperf-1.6.3-win32>iperf -s -p 10000 -i 1 -w 58K 
------------------------------------------------------------ 
Server listening on TCP port 10000 
TCP window size:  8.0 KByte (WARNING: requested 58.0 KByte) 
------------------------------------------------------------ 

D:\installers\iperf-1.6.3-win32>iperf -s -p 10000 -i 1 -w 16K 
------------------------------------------------------------ 
Server listening on TCP port 10000 
TCP window size:  8.0 KByte (WARNING: requested 16.0 KByte) 
------------------------------------------------------------ 

D:\installers\iperf-1.6.3-win32>iperf -s -p 10000 -i 1 -w 64K 
------------------------------------------------------------ 
Server listening on TCP port 10000 
TCP window size:  8.0 KByte (WARNING: requested 64.0 KByte) 
------------------------------------------------------------ 

============================================================ 

Whenever I start a UDP test with the Windows machine as the server and the
Solaris machine as a client, I get the following error

D:\installers\iperf-1.6.3-win32>iperf -s -p 10000 -u -i 1 
------------------------------------------------------------ 
Server listening on UDP port 10000 
Receiving 1470 byte datagrams 
UDP buffer size:  8.0 KByte (default) 
------------------------------------------------------------ 
[136] local 192.168.1.69 port 10000 connected with 172.16.1.1 port 37270 
[ ID] Interval       Transfer     Bandwidth       Jitter   Lost/Total
Datagrams 
[136]  0.0- 1.0 sec   128 KBytes   1.0 Mbits/sec  2.149 ms 1937339228/   89
(2.2 
e+009%) 
[136]  1.0- 2.0 sec   128 KBytes   1.0 Mbits/sec  2.356 ms    0/   89 (0%) 
[136]  2.0- 3.0 sec   129 KBytes   1.1 Mbits/sec  1.967 ms    0/   90 (0%) 
[136]  3.0- 4.0 sec   128 KBytes   1.0 Mbits/sec  2.096 ms    0/   89 (0%) 
[136]  4.0- 5.0 sec   128 KBytes   1.0 Mbits/sec  2.287 ms    0/   89 (0%) 
[136]  5.0- 6.0 sec   129 KBytes   1.1 Mbits/sec  1.915 ms    0/   90 (0%) 
[136]  6.0- 7.0 sec   126 KBytes   1.0 Mbits/sec  2.123 ms    0/   88 (0%) 
[136]  7.0- 8.0 sec   128 KBytes   1.0 Mbits/sec  2.288 ms    0/   89 (0%) 
[136]  8.0- 9.0 sec   129 KBytes   1.1 Mbits/sec  1.918 ms    0/   90 (0%) 
[136]  9.0-10.0 sec   128 KBytes   1.0 Mbits/sec  2.068 ms    0/   89 (0%) 
[136]  0.0-10.0 sec   1.3 MBytes   1.0 Mbits/sec  2.032 ms    0/  893 (0%) 

Why the strange packet loss number at the beginning of the test? 

==================================================== 

Finally, why can't iperf be consistent reporting intervals. For example the
following run generates a warning. This happens most frequently on slower
connections where the performance may drop close to zero for a given report
interval.

[188]  2.2- 3.1 sec  64.0 KBytes   646 Kbits/sec 
[188]  3.1- 4.3 sec  72.0 KBytes   491 Kbits/sec 
[188]  4.3- 5.1 sec  48.0 KBytes   491 Kbits/sec 
[188]  5.1- 6.8 sec  94.6 KBytes   440 Kbits/sec 
[188]  6.8- 7.2 sec [Warning] Skipping report, interval too small 
[188]  6.8- 8.0 sec   113 KBytes   760 Kbits/sec 
[188]  8.0- 9.2 sec  72.0 KBytes   491 Kbits/sec 
[188]  9.2-10.0 sec  48.0 KBytes   491 Kbits/sec 
[188]  0.0-10.1 sec   704 KBytes   574 Kbits/sec 

============================================================= 
And one feature request 
Can the reporting accuracy be increased to at least two decimal places for
reports in MBits 

We do a lot of testing on links that run 1-2 Mbits and 1 decimal place is
not enough. 


Thanks 
Paul 

--
Mitch Kutzko | mitch --at-- dast.nlanr.net | mitch --at-- ncsa.uiuc.edu | 217-333-1199
http://hobbes.ncsa.uiuc.edu/



Other Mailing lists | Author Index | Date Index | Subject Index | Thread Index