Re: [PATCH] for 2.2.14 to improve SpecWeb96 performance

From: Takashi Richard Horikawa (horikawa@ccm.cl.nec.co.jp)
Date: Fri Mar 31 2000 - 23:16:55 EST

  • Next message: Albert D. Cahalan: "ip_local_port_range"

    In message <38E4D8E7.9A73CF84@alumni.caltech.edu>, Dan Kegel writes:
    >Takashi Richard Horikawa (horikawa@ccm.cl.nec.co.jp) wrote:
    >This is a benchmarking issue that happens only when you try
    >to simulate too many clients with a single client machine, right?

    Exactly. I think that hitting TIME_WAIT'er by a connection request is
    quite rare in real-world WEB environment, in which many clients
    having individual IP addresses issue connection requests.

    I would like to point out that the problem that I found occurs only when
    a connection request hits a TIME_WAIT'er socket whose sock_readers
    is not zero. I summarize my finding in the following.

    Linux Connection Value of Result
    Version request sk->sock_readers (Reply the request?)
    ----------------------------------------------------------------
    2.2.14 Not hit TIME_WAIT'er - OK
    2.2.14 Hits TIME_WAIT'er zero OK
    2.2.14 Hits TIME_WAIT'er non-zero NG (Does not
                                                       reply the request)

    It may be odd that a TIME_WAIT'er has non-zero sockreaders.

    >(I've added a note about this at
    > http://www.kegel.com/mindcraft_redux.html#tcp3 )

    Thank you for introducing my article.

    >Did you try to tell 2.2.14 to use a larger port range, e.g.
    > # echo 1024 65535 > /proc/sys/net/ipv4/ip_local_port_range
    > # cat /proc/sys/net/ipv4/ip_local_port_range
    > 1024 65535
    >on your unpatched runs?

    Thank you for your advice. I'll try.
    --------
                                              name: Takashi Richard Horikawa
                                                    (NEC C&C Research Lab.s)
                                        E-mail: horikawa@ccm.cl.nec.co.jp
                                              Tel.: +81-44-856-2079

    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.rutgers.edu
    Please read the FAQ at http://www.tux.org/lkml/



    This archive was generated by hypermail 2b29 : Fri Mar 31 2000 - 23:16:50 EST