FW: NTPd code modification. - NTP

This is a discussion on FW: NTPd code modification. - NTP ; Hi All, I would like to introduce myself as Suhas.P.B , a Software Engineer from Bangalore, India. I have a special requirement that , ntp daemon should not modify the time automatically , instead it should inform a application ( ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: FW: NTPd code modification.

  1. FW: NTPd code modification.

    Hi All,

    I would like to introduce myself as Suhas.P.B , a Software Engineer from
    Bangalore, India.

    I have a special requirement that , ntp daemon should not modify the time
    automatically
    , instead it should inform a application ( some critical application ) that
    is running, the
    application should take care of modifying the time .

    The requirement is because my application runs timers, if the time is
    modified without notice
    to the application, the timers get expired and lead to a chaos.

    So , my initial doubts are ,

    Can i do this , by modifying the NTPd source code, if so any material i can
    liik into .
    Which part of the NTP code sets the system time .

    Another small requirement is that , Can i restrict the NTP to adjust the
    time, iff and only if
    the time difference crosses some limit with respect to the server

    Awaiting for your responses ,


    Thanks in advance,
    Suhas.P.B
    Lucent Technologies
    ST-1, Hosur Road, Bangalore - 560095.
    Tel: +91-80-51191355
    Mobile: +91-9886081128

    Lucent Technologies, Bell Labs Innovations



    <http://promos.hotbar.com/promos/prom...&RAND=90171&pa
    rtner=hbtools> Upgrade Your Email - Click here!

    _______________________________________________
    questions mailing list
    questions@lists.ntp.isc.org
    https://lists.ntp.isc.org/mailman/listinfo/questions


  2. Re: NTPd code modification.

    "P B, Suhas Suhas" wrote in message
    news:3BE48DD0EC7D3948BC183931D9150C2104387944@ii00 15exch001u.iprc.lucent.com...
    [...]
    > I have a special requirement that , ntp daemon should not modify the
    > time automatically , instead it should inform a application ( some
    > critical application ) that is running, the application should take
    > care of modifying the time .
    >
    > The requirement is because my application runs timers, if the time
    > is modified without notice to the application, the timers get expired
    > and lead to a chaos.


    I take it your application is that critical one.

    Good news! NTP doesn't work the way you think. Rather than adjusting
    the 'hands' of the clock directly, it adjusts the speed at which they
    move. That means that the timers never notice.

    Groetjes,
    Maarten Wiltink



  3. Re: NTPd code modification.

    Maarten Wiltink wrote:

    > "P B, Suhas Suhas" wrote in message
    > news:3BE48DD0EC7D3948BC183931D9150C2104387944@ii00 15exch001u.iprc.lucent.com...
    > [...]
    >
    >>I have a special requirement that , ntp daemon should not modify the
    >>time automatically , instead it should inform a application ( some
    >>critical application ) that is running, the application should take
    >>care of modifying the time .
    >>
    >>The requirement is because my application runs timers, if the time
    >>is modified without notice to the application, the timers get expired
    >>and lead to a chaos.

    >
    >
    > I take it your application is that critical one.
    >
    > Good news! NTP doesn't work the way you think. Rather than adjusting
    > the 'hands' of the clock directly, it adjusts the speed at which they
    > move. That means that the timers never notice.


    I think that would depend on the interval being timed and the precision
    of the local clock. Very short intervals could be significantly
    affected by nptd's operations.

  4. Re: NTPd code modification.

    "Richard B. Gilbert" wrote in message
    news:us2dnW2905coDj_ZnZ2dnUVZ_r-dnZ2d@comcast.com...
    > Maarten Wiltink wrote:


    >> [...] timers never notice.

    >
    > I think that would depend on the interval being timed and the precision
    > of the local clock. Very short intervals could be significantly
    > affected by nptd's operations.


    Yes, but.

    Timers with periods close to the OS's clock resolution are going to
    give bad results no matter what you do. Quantisation noise leading to
    elevated jitter is the least one may expect.

    I am assuming that (application) timer ticks are triggered from (OS)
    clock ticks, so that timer ticks can happen only at even multiples of
    the clock tick interval. Any other solution seems to me to require
    extensive spinning inappropriate for a general-purpose OS. Hardware-
    assisted solutions are thinkable but would pose their own problems of
    interaction with NTP.

    Clock interpolation using cycle counters should not affect these
    problems, or their solutions. The current actual tick length can and
    should be used instead of the nominal tick length in the interpolation.

    Groetjes,
    Maarten Wiltink



+ Reply to Thread