[ntp:questions] root vs peer dispersion?

David Woolley david at ex.djwhome.demon.co.uk.invalid
Mon May 4 17:42:38 UTC 2009


Towli wrote:

> 
> why is the root dispertion not "0" when the Dcf77-clock is sitting on 
> port 20 on the switch? is 5.45 the time the Dcf77-signal takes to travel 
> from Frankfurt to Denmark where i am located? What then does the peer 
> dispersion time of 2.84 msec tell me?

Because:

- the DCF 77 time wasn't read immediately before you requested the time 
from NTP, so the time could have drifted since then.
- I'm not sure, but I think that the best sample might not be the most 
recent;
- there is an uncertainty in reading the clock due to the finite 
resolution of the clock;
- I think there may be an element of the historic variability thrown in;
- using too low a root dispersion means the error band is very small, so 
it only needs a small error to make it impossible to get overlapping 
error bands.




More information about the questions mailing list