Date: Thu, 7 Feb 2002 12:55:10 +0000 (GMT) From: Mike Silbersack <silby@silby.com> To: murthy kn <knmurthy30@hotmail.com> Cc: freebsd-net@FreeBSD.ORG, <tinguely@web.cs.ndsu.nodak.edu> Subject: Re: Duplicate Acks and Fast Retransmit Message-ID: <20020207123715.H3102-100000@patrocles.silby.com> In-Reply-To: <F15115fyy6EcpOSvjbe0000a4d6@hotmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 7 Feb 2002, murthy kn wrote: > 1. In my understanding, this should make the "sending" TCP get into Fast > Retransmit mode each time it receives 2 identical acks (or 1 duplicate ack). > However, I am not seeing this behaviour and would like > to get your advice on if I am missing something. One of the reasons this appears to be the case is because I don't see duplicate acks (be careful to distinguish between duplicate acks and window updates.) In fact, I don't really see much in the way of packet reordering. > 3. I am also interested in knowing, in general, as you have > indicated, why we cannot draw inferences from "netstat". Well, I suppose that you _can_ draw inferences, but they're not going to be accurate. If you're doing any serious network research, you need to trace it exactly using tcpdump or some similar tool. You cannot tune a network stack through statistics alone. > For eg, in my case, once I start packet capture on the > sender machine, I will not see the duplicate acks or > reordering - this has mostly to do with the fact that > reordering also depends on the load u put (amongst other things). > > Since I have a virtual interface, I do not know if it is > possible to use another machine in a passive monitoring mode > to capture the packets on the "virtual" interface of this > mahine - as there are 2 wires and not 1. (Apart from using > 2 other machines and kind of merging the 2 dumps, is there any other way to > capture packets/analyse TCP behaviour in this kind of scenarios??) I'm not sure what's the best way to configure your test setup. Bill Fenner might have some suggestions. > Since I HAD to run the capture on the reciver, tcpdump reported > drops and some packets seem to have been dropped... You tend to drop less packets if you use tcpdump's -w option, then use -r later. Mike "Silby" Silbersack To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020207123715.H3102-100000>