Communication problem in UDP

That is difficult, as your sniffer might also lose packets unless the line speed is fairly slow (in which case the client should not lose packets either), or unless the sniffer is well designed for sniffing.

There was a rant on this topic earlier this year,

Only with the cooperation of both sides to change the application protocol.

If you have control over the sending machine but cannot change the receiver, then you could put information into the IP Options portion of the IP header, such as option 68 "Time Stamp" or option 136 "Stream ID". "Time Stamp" would seem to be a relatively good fit: for your purposes, increment it by 1 for each packet instead of trying to use some kind of real time measurement.

Sound advice.

If you suspect lost packets he very first thing to do is to eliminate a duplex missmatch or other physical layer errors from the investigation.

Do a show int and check for input and outpur errors.

If you wish post the output here.

I seem to recall that IP packets have a 16 bit datagram number (I have just called it). "Identification ( 16-bit number which together with the source address uniquely identifies this packet - used during reassembly of fragmented datagrams)

If you are lucky your system will be incrementing it for each packet.

Hi all,

Thanks a lot Walter for your answer. You are every time present to help and to give answer to the questions. Many thanks for this great help.

Ok for the possibility to the sniffer to lost packet. I am using EtherReal, i will try to use others and compare. Well.

I cannot change the communication btw client and server, so i cannot insert TAG to the packet transmited by the server.

But, "anybody43" sended an answer which seem to be well for me. I will answer in fiew minutes to the same post... check the message please !! :)

Thanks a lot

Best Regards Rahan

"Walter Roberson" a écrit dans le message de news:OApAg.314824$IK3.7384@pd7tw1no...

ThankYou very much anybody43 for your answer !!

I confirm that i don't have any errors on the interface of my Cisco switch (sh int | include errors).

I confirm that the "identification" of the UDP packet is incrementing. For exemple :

... > 0xe002 > 0xe003 > 0xe004 > 0xe005 > 0xe006 > 0xe007 > 0xe008 > 0xe009

0xe012 > 0xe013 > 0xe014 > 0xe015 > 0xe016 > 0xe017 > 0xe018 > 0xe019 > 0xe01a > 0xe01b > 0xe01c > ...

Can you confirm please that this "identification" is unique and i can check it to be sure if some UDP packet are lost ?

In the same time, under client Windows (XP sp2), when i run "netstat -es" i have a lot of "Received Address Errors". This counter is incrementing and i don't know why !! By analizing my sniffer, i don't have any problem around Address of the sender or receiver !!

In the same client, i stopped "flow control", i stopped "checksum offload" and i disabled QOS on the NIC and the counter of "Received Address Errors" is always incrementing !! I changed the cable and the network card (broadcom to 3com) and i experimented the same problem.

I reminder you that data sended by the server to the client is in multicast (udp) and i am sure that the client is always in the multicast group and never leave this group and the client always recevive the data to the multicast group.

Any idea about this question please ?

Thank You very much

Best Regards Rahan

a écrit dans le message de news: snipped-for-privacy@m73g2000cwd.googlegroups.com...