[3085] in linux-net channel archive

home help back first fref pref prev next nref lref last post

Re: Tcpdump interpretation, again

daemon@ATHENA.MIT.EDU (athan@mersinet.co.uk)
Sat Jun 1 20:07:25 1996

To: linux-net@vger.rutgers.edu
Date: 	1 Jun 1996 23:24:58 GMT
From: athan@mersinet.co.uk

In article <31AF5583.4F8AEF5A@ix.netcom.com>,
Robert Wuest <rwuest@ix.NETcom.COM> wrote:
>I don't want to appear too stupid, but I really am trying
>to figure this out.
  I know that feeling...
>I'm running (I remember, Michael :-) ) Linux 1.99.9, ppp-2.20f,
>Machine is a P100 w/48M, and a hodgepodge of drives, SCSI, 
>EIDE (if I could find it working cheap, it's in my box!). 
>Downloading with Netscape NavigatorTM Gold Version 3.0b4.
>
>I was ftping a file, and it was going pretty slow (460 bps on a 14.4
>connection) , so I took a gander w/ tcpdump and found this.
[tcpdump output snipped]
   Hmmm, well, I had a similar problem, but with the transfer going the
other direction. My transfers INTO my machine were ok, OUT was way slow
(~15Bytes/s on 14.4kbps modem OUT, ~1.1KB/s IN). Turned out that at least
a partial solution was to set the speed between the modem and serial
port to 38400 or 19200, rather than the 115.2kbps I'd been using DESPITE
the fact the modem has a 16550A UART...

>I don't get it! Lot's of wasted bandwidth here with some of these packets being
>sent three or four times.  Is princeton's server messing up.  Surely, this isn't
>right.

  The tcpdump IS very strange, repeatedly ACK'ing a packet that was
shown to be received... If you have your modem/serial port speed above
19200 try dropping it and seeing what happens...

-Neil
-- 
**************************************************************************
* Neil Soveran-Charley, System Administrator, Mersinet Internet Services *
* Email: N.P.Soveran-Charley@mersinet.co.uk                              *
**************************************************************************


home help back first fref pref prev next nref lref last post