[ntp:questions] Re: NTP stepping issue

Philip Homburg philip at pch.home.cs.vu.nl
Mon Oct 25 17:24:44 UTC 2004


In article <417D2F21.6090203 at udel.edu>, David L. Mills <mills at udel.edu> wrote:
>We are talking right past one another. First, roundtrip delays can 
>indeed exceed 128 ms and even much more, as with the Mars Internet 
>simulations reported on the web. That has nothing to do with jitter. 
>Jitter can exceed 128 ms on occasion, but the clock state machine 
>mitigates that. In addition, the huff-n'-puff scheme is a crude but 
>effective remedy for assymetric delays in at lease one common case.

I think there are plenty of documented cases whether people see jumps of
between 128 ms and, say, 10 seconds.

>The simulator code IS the actual running code includeing all the 
>algorithms anmd the clock discipline algorithm in particular - no change 
>other than to simulate network jitter and oscillator wander and the 
>kernel clock itself. There is no "interface" as such; the only system 
>calls are to adjtime() and settimeofday(), which are indeed simulated.

Of course there is an interface. My code doesn't use adjtime or settimeofday
(I created a kernel interface that fits my ideas of how time synchronization
should be done).

Apart from that, the clock discipline algorithm also has to be able to
get a new timestamp from the reference clock. And the algorithm under
test should not just call sleep because that would make it impossible
to do a discrete event simulation, so there has to be a way to schedule a
timeout event, etc.

A simple nm on ntp_loopfilter.o should make it clear that the actual
interface is quite complex.

Anyhow, I would like to stimulate people to experiment with other algorithms
than the ones used in the NTP implementation and compare (in vivo) how
they behave. Computers support a much wider range of synchronization
algorithms than can be implemented (effectively) in analog electronics
(or other analog systems). 


-- 
This Monk had first gone wrong when it was [...] cross-connected to a video
recorder that was watching eleven TV channels simultaneously, [...] The video
recorder only had to watch them, of course. It didn't have to believe them all
as well. This is why instruction manuals are so important    -- Douglas Adams



More information about the questions mailing list