Hello again,

Since the last round of posts, I have rejigged my test setup and started using 
Wireshark to snoop on the network traffic between my laptop client and the 
server.

The two are now connected directly via Gigabit ethernet cable without any 
intervening switches. I have also tried another cable just in case - no 
difference.

The client and server IPs are 192.168.0.67 and 192.168.0.100 respectively. As 
before I am using Iperf to generate traffic and measure the bandwidth. I also 
have a SSH session open.

I start Wireshark on the client and and invoke the test (also on the client). 
This sends roughly 1MB of data to the server.

aeon:~ andrewharvey$ iperf -c 192.168.0.100 -fM -t.05
------------------------------------------------------------
Client connecting to 192.168.0.100, TCP port 5001
TCP window size: 0.13 MByte (default)
------------------------------------------------------------
[  3] local 192.168.0.67 port 53160 connected with 192.168.0.100 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0- 0.1 sec  0.94 MBytes  18.7 MBytes/sec


As soon as the test ends I stop capturing. The resulting log file is attached 
to this message.

The results are interesting...

First of all, Wireshark flags all the packets sent from the client as having 
the wrong header checksum. This is because checksumming is offloaded onto 
hardware, and Wireshark is intercepting the outgoing packets before they get 
there. So we can safely ignore that issue.

However, the server appears to be acknowledging packets that haven't yet being 
sent:

241     0.030294        192.168.0.67    192.168.0.100   TCP     [TCP Previous 
segment lost] 53160 > commplex-link [PSH, ACK] Seq=310849 Ack=1 Win=524280 
Len=472 TSV=382288475 TSER=972369

One possible explanation is that Wireshark simply isn't capturing all the 
outgoing packets, and every now and then misses a few.

Wireshark also reports [TCP Previous segment lost] errors. For example:

364     0.038842        192.168.0.67    192.168.0.100   TCP     [TCP Previous 
segment lost] 53160 > commplex-link [ACK] Seq=507929 Ack=1 Win=524280 Len=1448 
TSV=382288475 TSER=972370

This may be related to the above ACK problem.

Additional strangeness occurs at the very end after the Iperf test when it 
attempts to close the connection. In the attached capture there are eight 
identical [FIN, ACK] packets and nine duplicate ACKS from the server!

I would be grateful if someone could give me their informed opinion as to 
whether there is anything suspicious going on here.

Thanks,

Andrew
-- 
This message posted from opensolaris.org
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to