[ntp:questions] NTP setup in time sensitive environment

NVAP nvap23 at gmail.com
Thu Apr 23 20:32:47 UTC 2015


Thank you for your reply.

1) The windows boxes may be rebooted by a customer few times a day and so
we do not have any control.
2) I could try removing burst and increase the min and maxpoll, like you
suggested.
We wanted the NTP client to trust the Server as soon as possible for time
synchronizations and hence been using minpoll = 3.
3) If PC_NTP was down for any reason, it might be acceptable for the system
time to be off.
The main intention is that PC_V, PC_A and PC_B are in time sync with
minimal to no time offset.



On Wed, Apr 22, 2015 at 5:54 PM, Charles Swiger <cswiger at mac.com> wrote:

> On Apr 22, 2015, at 3:24 PM, NVAP <nvap23 at gmail.com> wrote:
>
> *Questions:*
> 1) How can we assure that the sync will consistently result in +/- 5ms
> offset?
>
>
> If you weren't rebooting the Windows boxes potentially several times a day,
> Windows might do better.  I've seen RRD graphs of stratum-2 Win 7 boxes
> that meet your tolerance-- for example:
>
>    http://www.satsignal.eu/mrtg/performance_ntp.php
>
> 2) This seems to defeat the whole purpose of synching if the offset cannot
> be maintained!
> Is there a way to force the system offset to remain within +/-5ms?
>
>
> I'm not convinced that tuning the clients to use burst and minpoll 3 /
> maxpoll 6 is helping.
> Wasn't there a bug in Windows where very small offsets?  Ah, yes:
>
>    https://support.microsoft.com/en-us/kb/2537623
>
> Get rid of burst and try minpoll 6 maxpoll 8.
>
> 3) Can we live with just 1 time server PC_NTP and PC_V, PC_A and PC_B will
> closely following that time no matter?
>
>
> Yes.  Of course, if something happens to PC_NTP, it will become a single
> point of failure.
>
> Regards,
> --
> -Chuck
>
>


More information about the questions mailing list