[time-nuts] Time illiterate query about feasibility of determining source clock quality based on NTP jitter/disp and offset
Saku Ytti
saku at ytti.fi
Sat Nov 28 13:27:35 UTC 2009
cisco#show ntp associations
address ref clock st when poll reach
delay offset disp
*~meinberg-dhq .PPS. 1 13 256 377 2.6
1.53 3.7
+~meinberg-lq .PPS. 1 131 256 377 2.1
1.83 1.4
+~randomPC1 .GPS. 1 29 256 377 17.0
1.35 0.3
+~randomPC2 randomPC3 2 45 256 377 17.6 1.17 0.3
+~randomPC3 .GPS. 1 41 256 377 17.0
1.52 0.4
* master (synced), # master (unsynced), + selected, - candidate, ~ configured
I see similar result all around, meinbergs which are lot closer, have
poorer offset
and disp.
Are there any conclusions I can draw from this?
Funny thing is that cisco iscloser to randomPC2 than its source random PC3,
so I think just based on that, I can determine I can't use these values to
determine quality of source clock?
But what would then explain how disp/jitter and offset are consistently better
towards the randomPC's which are considerably further down the network?
With basically no clue about time, I have two uneducated guesses.
1) meinberg is embedded hardware, software processing is slow in it
compared to randomPC which runs software much faster, which is
much more visible in NTP than actual clock quality
or
2) cisco and randomPC have similar clock hardware, something bulk
most hardware use, these clocks skew to same direction at relatively
similar rate, making the cisco and randomPC's having time quite close
to each other, both being bit further away from better clock.
Thanks,
--
++ytti
More information about the Time-nuts_lists.febo.com
mailing list