[ntp:questions] strange behaviour of ntp peerstats entries.

root root at physics.ubc.ca
Fri Jan 25 09:25:24 UTC 2008

OK, having looked at the code, I see what it is doing. It essentially 
takes the measurement with the shortest delay in the past 8 measurements.
 If that measurement happens to be the current one, it is actually used 
in the clock loop. (Yes, I know this characterisation of the clock filter is cruder
than reality).

This makes the smaller variance of chrony even more impressive, since in my
 tests, evey single measurement of offset was used to calculate the variance 
in the case of chrony, while
for ntp, only those "smallest" values as reported in peerstats were used. This also 
explains why the roundtrip variance in ntp was so much better than chrony's. 

Unruh <unruh-spam at physics.ubc.ca> writes:

>Having been testing ntp, I was looking at the peerstats file. Here is a
>sequence of such measurements.

>54489 29187.448 9414 0.005089025 0.016225398 0.067088246 0.001122070
>54489 29269.446 9614 0.003231929 0.015419416 0.009606450 0.001005171
>54489 29316.445 9414 0.005089025 0.016225398 0.067796410 0.000929591
>54489 29396.445 9614 0.002083530 0.015706261 0.007395887 0.000386142
>54489 29443.447 9414 0.005089025 0.016225398 0.067421760 0.001041993
>54489 29523.443 9614 0.002083530 0.015706261 0.009098914 0.000252795

>I thought that there was an entry in this file for each packet exchange
>with the server. But notice the entries at 29396.445 and 29523.443 seconds.
>The offsets are completely identical. There is no way that this could be
>true if this were really the outcome of a measurement. There is just no way
>that you would get two entries the same to 7 decimal places ( well, it is
>really only 4 since the machines have a resolution of a microsecond). And
>there are a number of these. This simply cannot be the offset as obtained
>from a measurement. 
>What do these entries mean? How can the two offsets be identical?

>Note that I was watching successive output from ntpq -p at the same time.
>and the "when" entry went from 122 to 34-- ie indicated that a measurement
>was made-- and the offset did not change at all. Just before this in the
>peerstats file there were 7 entries with identical offsets and identical
>roudtrip delays.  What is happening here? What does the "offset" and
>"roundtrip delay" mean if not the measured offset (t4-t3+t1-t2)/2 and
>roundtrip (t4-t1) on each measurement?

More information about the questions mailing list