transparent tcp improvements

From: Karl Jonas ([email protected])
Date: Mon Apr 15 2002 - 04:10:35 EDT

  • Next message: Mike Jagdis: "Re: transparent tcp improvements"

    Dear Experts,

    I browsed through your mailing list archive, looked at a couple
    of referenced docs, looking for the following solution to the
    long-delay problem:

    having a mobile device attached to a long-delay wireless link,
    I would like to improve tcp's performance by sending early ACKs
    from a box at the 'end' of the terrestrial network towards the
    tcp sender in the terrestrial network:

    mobile -- (long delay wireless link) -- tcp box -- (terrestrial nw) --
    sender

    The idea is that tcp's slow start / congestion avoidance mechanism
    will handle the situation in the terrestrial nw correctly, and
    transmission will not suffer from the long delay link.

    The 'tcp box' must monitor tcp sesssions, generate acks towards the
    sender (possibly drop acks from the mobile?), and of course buffer
    tcp data that it has ack'ed to the sender, but which has not been
    ack'ed by the mobile yet.

    Since this is a very simple idea, I am sure that it has been
    implemented and evaluated.

    Please send me any reference to such an approach (I could not find one
    yet).

    Note that I am not interested in split-tcp, snoop, i-tcp or similar
    approaches that are non-transparent to the sender/receiver, or that
    need a modification on the mobile terminal side of the wireless link.

    Thanks for your help (please reply (also) to my address [email protected])

    Karl

    --
    Prof. Dr. Karl Jonas            EMail: [email protected]
    

    University of Applied Science FhG FOKUS CC SATCom Grantham-Allee 20 Schloss Birlinghoven 53757 Sankt Augustin D-53754 Sankt Augustin

    Tel: +49 (0)2241 865244 +49 (0)2241 141599 Fax: +49 (0)2241 141050 WWW: www.gmd.de/people/Karl.Jonas/




    This archive was generated by hypermail 2b29 : Mon Apr 15 2002 - 04:17:25 EDT