Re: Trying to understand strange behavior
I would also like to point out here the strange (large) numbers
you someimes get for the first report of lost datagrams / outoforder
datagrams count, with v1.6.5 on
win2k. It looks like some variables (eg: mPLastOutOfOrder) were not
initialized to 0 in function PerfSocket::InitTransfer (atleast in the
source code distribution). When I added the code, I no longer got the
erroneous large counts.
Vinay.
On Thu, 5 Jun 2003, paul shields wrote:
> I am seeing some strange behavior when using iperf in UDP mode from a Sun client to a Windows 2000 server (v1.6.5).
>
> Here is the server log
> C:\>iperf -s -u -l 1470 -i 5 -B 172.27.2.196
> ------------------------------------------------------------
> Server listening on UDP port 5001
> Binding to local address 172.27.2.196
> Receiving 1470 byte datagrams
> UDP buffer size: 8.00 KByte (default)
> ------------------------------------------------------------
> [136] local 172.27.2.196 port 5001 connected with 172.27.2.194 port 34205
> [ ID] Interval Transfer Bandwidth Jitter Lost/Total Datagrams
> [136] 0.0- 5.0 sec 581 KBytes 951 Kbits/sec 6.208 ms 1735357008/ 405 (4.3
> e+008%)
> [136] 5.0-10.0 sec 613 KBytes 1.00 Mbits/sec 3.922 ms 77/ 504 (15%)
> [136] 10.0-15.0 sec 614 KBytes 1.01 Mbits/sec 4.234 ms 442/ 870 (51%)
> [136] 15.0-20.0 sec 614 KBytes 1.01 Mbits/sec 5.024 ms 445/ 873 (51%)
> [136] 20.0-25.0 sec 614 KBytes 1.01 Mbits/sec 4.625 ms 442/ 870 (51%)
> [136] 25.0-30.0 sec 613 KBytes 1.00 Mbits/sec 4.501 ms 442/ 869 (51%)
> [136] 0.0-34.8 sec 4.14 MBytes 998 Kbits/sec 4.083 ms 2273/ 5226 (43%)
> recvfrom failed: Connection reset by peer
>
> C:\>
>
> And the line on the client used to start the test is:
> iperf -c 172.27.2.196 -u -b 2048K -l 1470 -t 30
>
> Why is the bandwidth number specified not delivered from the start of the test. It seems like it takes 10 seconds before the packet rate matches that specified in the startup.
>
> Paul
>
>