[ntp:questions] ntpd and database servers

David Woolley david at ex.djwhome.demon.invalid
Thu Jan 28 23:11:44 UTC 2010


Evandro Menezes wrote:
> Doesn't NTP know UTC via the root dispersion?  If so, then instead of
> giving out its time, it could give out UTC.  I'm sure that there may
> be more to it, but just a thought.

This doesn't make sense to me.

If the root time is UTC.  NTP time will be approximately UTC.  The error 
should be less than the root distance, but one cannot be more certain 
that that.

Root dispersion is one component of that error budget.  It is a near 
worst case estimate of how much the time may have varied from root time 
as the result of uncompensated clock frequency error, and measurement 
resolution limitations.  The effect of this term is to create an error 
of either sign with possible values including zero.

NTP does not report the time which the reference clock last returned, 
but rather a smoothed version of this, which most importantly in this 
context, has been aged by its best estimate of 1 second for every second 
since it was read.  Dispersion is the uncertainty in the accuracy of 
that aging rate.




More information about the questions mailing list