Adrian von Bidder wrote:
What are your timing requirements?
For me (and I claim for most others), precision is just not an issue - a
single PC needs to be synced to within a few seconds of UTC, and ntp does
far better than that. A cluster with shared filesystems etc. needs to be
synced internally to not confuse make etc. with file timestamps, and to
satisfy kerberos etc. - but then you'd sync the cluster internally anyway,
so this becomes a non-issue again.
True, but since I've got a server in the pool I'd rather be as reliable
as possible... If using a remote stratum 1 is going to mean I'm a
couple of seconds out then it's not worth it - I'd rather not introduce
any *more* error than is already there.
reliability is an issue - and I guess stratum 1 and stratum 2 is not the
correct question here. Stratum 1 run by some hobbyist on a DSL line won't
be reliable, but stratum 3 used as main timeserver at a big university will
usually be. (time[123].unizh.ch, for example)
It comes down to.. which is more accurate:
A stratum 2 in the UK:
veracity.mcc.ac 193.63.105.18 2 u 354 1024 377 23.520 0.023
0.037
Or a stratum 1 in Germany:
sombrero.cs.tu- .PPS. 1 u 500 1024 377 50.280 -0.473
0.202
OTOH you're saying this is more accurate (actually a stratum 2 not 3):
rosehip.exnet.c 192.43.244.18 2 u 7 64 37 37.404 -5.812
1.827
..which means I really don't understand how to interpret the figures.
ntpd itself says that the stratum 1 is the most accurate, and picks it
as a timesource.
My reading of the figures says that the UK source is most accurate,
having the lowest jitter, delay and offset.
Now I'm really confused!! Is there a document that explains how to
interpret these figures?
Tony
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers
