Max polling time - NTP

This is a discussion on Max polling time - NTP ; I'm going to set Minpoll and Maxpoll time in my ntpd.conf for my NTP internal server (MASTER) which must be synchronized to external servers (stratum 1 or 2), because I think 64s is too low so i want to increase ...

+ Reply to Thread
Results 1 to 6 of 6

Thread: Max polling time

  1. Max polling time

    I'm going to set Minpoll and Maxpoll time in my ntpd.conf for my NTP
    internal server (MASTER) which must be synchronized to external
    servers (stratum 1 or 2), because I think 64s is too low so i want to
    increase it.
    I have another 2 internal time servers (A, B) which must be
    synchronized to my MASTER server.
    Clients will use A and B servers.

    Which time do you suggest me for A, B and MASTER servers ? I know that
    these values represent degree of 2.
    Maxpoll ?
    Minpoll ?
    Are These option supported in ntp 4.x in Fedora Linux ?


  2. Re: Max polling time

    In article <1173904589.823036.139010@e1g2000hsg.googlegroups.c om>,
    "RICCARDO" writes:
    >I'm going to set Minpoll and Maxpoll time in my ntpd.conf for my NTP
    >internal server (MASTER) which must be synchronized to external
    >servers (stratum 1 or 2), because I think 64s is too low so i want to
    >increase it.


    How long did you let your system run?

    NTP adjusts the polling interval. Usually it does the right thing.
    If the temperature and/or load on your system and/or the delays on
    your network connection are all stable, it should go up to 1024 seconds.
    It takes a while.

    --
    These are my opinions, not necessarily my employer's. I hate spam.


  3. Re: Max polling time

    My system is always run (server Master, server A , server B).
    I think ntpd polling time was 64s as default but you'are telling me
    that ntpd uses variable polling time according to environment and
    system status ?
    Hal Murray ha scritto:

    > In article <1173904589.823036.139010@e1g2000hsg.googlegroups.c om>,
    > "RICCARDO" writes:
    > >I'm going to set Minpoll and Maxpoll time in my ntpd.conf for my NTP
    > >internal server (MASTER) which must be synchronized to external
    > >servers (stratum 1 or 2), because I think 64s is too low so i want to
    > >increase it.

    >
    > How long did you let your system run?
    >
    > NTP adjusts the polling interval. Usually it does the right thing.
    > If the temperature and/or load on your system and/or the delays on
    > your network connection are all stable, it should go up to 1024 seconds.
    > It takes a while.
    >
    > --
    > These are my opinions, not necessarily my employer's. I hate spam.



  4. Re: Max polling time

    Riccardo,

    Yes, that's right. NTP uses adaptive algorithms to modify the polling
    interval.
    Usually it starts at 64s and then, at a rate depending on lots of
    factors,
    increases to 1024s (steady-state).

    For details, see slides 4 and 13 of
    http://www.eecis.udel.edu/~mills/dat...lock/clock.ppt
    and also slide 35 of
    http://www.eecis.udel.edu/~mills/dat.../flow/flow.ppt

    There's a lot of science in NTP and I would discourage you from
    modifying
    minpoll/maxpoll unless you have plenty of experience with NTP and
    a very good reason for doing it.

    Paul

    On 15 Mrz., 08:35, "RICCARDO" wrote:
    > My system is always run (server Master, server A , server B).
    > I think ntpd polling time was 64s as default but you'are telling me
    > that ntpd uses variable polling time according to environment and
    > system status ?
    > Hal Murray ha scritto:



  5. Re: Max polling time

    RICCARDO wrote:
    > I'm going to set Minpoll and Maxpoll time in my ntpd.conf for my NTP
    > internal server (MASTER) which must be synchronized to external
    > servers (stratum 1 or 2), because I think 64s is too low so i want to
    > increase it.
    > I have another 2 internal time servers (A, B) which must be
    > synchronized to my MASTER server.
    > Clients will use A and B servers.
    >
    > Which time do you suggest me for A, B and MASTER servers ? I know that
    > these values represent degree of 2.
    > Maxpoll ?
    > Minpoll ?
    > Are These option supported in ntp 4.x in Fedora Linux ?
    >


    It is generally a BAD IDEA to change the default values of MINPOLL and
    MAXPOLL. NTPD will change the polling interval to the value within
    those limits that is best suited to the existing situation. When ntpd
    is first started all the servers will be polled at intervals of MINPOLL
    seconds. As ntpd gains confidence that it knows what time it is and
    which are the best servers, it will increase that polling interval. If
    conditions (primarily network delay times) change, it will adjust the
    polling interval for best performance.


  6. Re: Max polling time

    Richard B. gilbert wrote:
    > RICCARDO wrote:
    >> I'm going to set Minpoll and Maxpoll time in my ntpd.conf for my NTP
    >> internal server (MASTER) which must be synchronized to external
    >> servers (stratum 1 or 2), because I think 64s is too low so i want to
    >> increase it.
    >> I have another 2 internal time servers (A, B) which must be
    >> synchronized to my MASTER server.
    >> Clients will use A and B servers.
    >>
    >> Which time do you suggest me for A, B and MASTER servers ? I know that
    >> these values represent degree of 2.
    >> Maxpoll ?
    >> Minpoll ?
    >> Are These option supported in ntp 4.x in Fedora Linux ?
    >>

    >
    > It is generally a BAD IDEA to change the default values of MINPOLL and
    > MAXPOLL. NTPD will change the polling interval to the value within
    > those limits that is best suited to the existing situation. When ntpd
    > is first started all the servers will be polled at intervals of MINPOLL
    > seconds. As ntpd gains confidence that it knows what time it is and
    > which are the best servers, it will increase that polling interval. If
    > conditions (primarily network delay times) change, it will adjust the
    > polling interval for best performance.


    I'd also like to point out that the increase in the polling interval is
    part of the engineering of the algorithms since it also acts as a
    dampener to wild swings that may occur as it tries to approach *best
    chime*. If you don't understand the algorithms you shouldn't touch the
    poll settings.

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


+ Reply to Thread