[chrony-dev] Re: [ntp:questions] Visualization of clock control

[Miroslav Lichvar <mlichvar@xxxxxxxxxx>]

On Thu, Jan 05, 2012 at 11:40:25AM +0800, Dennis Ferguson wrote:
> On 4 Jan, 2012, at 22:54 , Miroslav Lichvar wrote:
> > The simulations were done with a clock wandering at 1 ppb/s,
> > 10/100/1000us network jitter with exponential distribution and the NTP
> > clients were configured to use 64s polling interval.
> 
> That's pretty neat.  I think, however, that the clock wander of 1 ppb/s
> is about an order of magnitude too large for real life, at least for machines
> kept in an air conditioned room (and the behavior of clocks in machines
> subject to environmental variations probably can't be modeled by "wander" at
> all).  My measurements against precise hardware tended towards a value of
> 1ppb/10s, which is also consistent with the 10^-8/1000s which sometimes shows
> up on Allan variance plots (I think there's a square root relationship in there
> if the wander is a truly random walk).

I think the 1ppb/10s random walk wander corresponds to ~0.32ppb/1s.
The +0.5 slope in the variance plot intersecting 10^-8 at 1000s would
be ~0.6ppb/s wander.

I tried to model some thermal effects by adding a sine, triangle or
pulse wave to the clock frequency, but it seemed to me the effect it
had on the overall RMS time error was similar to just increasing the
wander. So instead of three or more parameters of the clock I set only
one. Sometimes I use even 10ppb/s wander, to simulate a machine with
varying CPU load and I think the results are not that different from
what I see on my desktop.

BTW, the simulator can be configured to read the clock frequency from
a file. If you have real data from a PPS refclock, you can use that
and see at what random walk wander will ntpd give similar results.

> The other difficulty with respect to real life may be modeling network jitter
> as exponential, since I believe the probability distribution for network delays
> is heavy-tailed (i.e. with extreme values way over-represented; this is a problem
> when using statistics which assume the underlying error distribution is gaussian).
> I don't know how to fix that, though.

I'd definitely be interested in a better model for network delays. I
guess we could try to make a collection of the ntp rawstats logs from
various network environments and see how the distribution looks like.

-- 
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.