[ntp:questions] Polling vs. Offset
unruh at invalid.ca
Sat Apr 27 16:48:35 UTC 2013
On 2013-04-27, Mischanko, Edward T <Edward.Mischanko at arcelormittal.com> wrote:
> Is there a way to configure ntpd so that when the offset exceeds 1
> millisecond polling will automatically be set to 64 seconds or less?
> If I understand correctly, polling currently is adjusted by the speed of
> change in offset. It takes forever for a large offset to be skewed at
> large polling time cycles when the drift is so small that it doesn't
> trigger a polling adjustment. Maybe additional code is needed to configure
> this function?
1ms used to be considered highly accurate time, and to have changed
procedure for that would have meant ntpd never settled down. Thus any
such policy would have to be dependent on the accuracy of the source it
is tracking and the accuracy of the clock on the computer, something ntp
has no idea about. ntpd is also designed to care more about accuracy in
drift rate than accuracy in offset ( although they are clearly related).
And longer polls are better for accurate drift rate for a procedure that
has no memory like ntpd. As Mills has repeatedly said, he has no
interest in ntpd's speed of convergence.
More information about the questions