Re: [chrony-dev] New chrony website

[Miroslav Lichvar <mlichvar@xxxxxxxxxx>]

On Tue, May 05, 2015 at 08:58:00AM -0700, Bill Unruh wrote:
> Looks good. I am surprized by the the tests-- at 1 ms ntp and chrony seem to
> be the same, with them differeing for all other tests. That is weird. Why 1ms?
> Also both other (ntp and openntpd) are flat below 1ms, and then take off. That
> looks strange. Test 1, chrony basically grows as the square root of the
> jitter. (100us* sqrt(jitter/100usec). Again, why 100us? The square root I
> can understand.

Here is a graph that probably shows this better (for ntpd look at the
green line, openntpd is not included):

https://mlichvar.fedorapeople.org/clknetsim/chrony_ntp/client_poll6-10_jitter_offset.png

I think the reason why at 1ms jitter chrony and ntp are so close is
that it's the point where the PLL time constant selected by ntpd is
closest to the optimal value. With smaller jitter the error is mostly
from clock wandering, in NTP terminology it could be said that Allan
intercept is too short for the PLL time constant.

BTW, I have added third test for intermittent connection to the
comparison page.

-- 
Miroslav Lichvar

-- 
To unsubscribe email chrony-dev-request@xxxxxxxxxxxxxxxxxxxx with "unsubscribe" in the subject.
For help email chrony-dev-request@xxxxxxxxxxxxxxxxxxxx with "help" in the subject.
Trouble?  Email listmaster@xxxxxxxxxxxxxxxxxxxx.