Re: question

From: Shawn Ostermann (sdo@picard.cs.ohiou.edu)
Date: 03/25/01

  • Next message: Bogdan Ghita: "Error in cwin estimation?" 

    Message-Id: <200103252025.PAA03369@picard.cs.ohiou.edu>
Date: Sun, 25 Mar 2001 15:25:06 -0500
From: Shawn Ostermann <sdo@picard.cs.ohiou.edu>
Subject: Re: question

    ------- Blind-Carbon-Copy

    To: Yong Xia <xiay@rpi.edu>
    In-Reply-To: <3ABD7BE0.9D183600@rpi.edu>
    Reply-To: ostermann@cs.ohiou.edu
    From: "Shawn Ostermann" <ostermann@cs.ohiou.edu>
    Subject: Re: question
    Date: Sun, 25 Mar 2001 15:25:06 -0500
    Sender: sdo@picard.cs.ohiou.edu

    >> I am wondering if there is something in the tcptrace or xplot need
    >> to be improved. btw, the graph of cwin.xpl is calculated from
    >> tcpdump data. but cwin is itself a kernel varible inside the
    >> OS. so you acctually don't dump it from tcpdump info, but compute
    >> it from that info, do you?

    Well, this is just a heuristic at best. tcptrace is inferring cwin
    from the UNacked data on the wire. That's normally a pretty good
    estimate. Of course, the only place you can really get that
    information is from the sending TCP's kernel tables. I've never seen
    a case where it's obviously wrong, though. Perhaps the sending TCP
    that you're looking at is doing it wrong.

    - --sdo

    ------- End of Blind-Carbon-Copy

    This archive was generated by hypermail 2b30 : 03/26/01 EST