[ntp:questions] NTP no internet connection

Unruh unruh-spam at physics.ubc.ca
Sun Feb 10 12:04:33 UTC 2008

"David J Taylor" <david-taylor at blueyonder.neither-this-bit.nor-this-bit.co.uk> writes:

>Unruh wrote:
>> ??? He is on a private network with absolutely no access to the
>> network. What the network time is is irrelevant.
>> I do not dispute that having a GPS receiver will make sure that that
>> server keeps very close ( a few usec) to "true time". But it will
>> make no difference to how well his various machines are synched to
>> each other.

>Do you have any measurements to show that is true with NTP?  My gut 
>feeling is that the more stable the reference source, the more closely its 
>slaves could be synched.

On the microsecond level yes. Ie, in his case ntp will have to correct for
both source drift and client drift-- two places where large changes( few
PPM) in the drift can occur and since ntp handles changes only slowly, both
will add extra noise onto the sync. But this is at the usec level,
certainly not the msec level.
Note that if he changes the time on his system by for example using date
and his wristwatch to reset the time on the server, it will take a while
for  the rest of the systems to come into sync ( a few hours) to the
original accuracy.
So yes, your gut feelings are correct to a certain extent. However, on the
level of accuracy he is proposing, it will make no difference.


More information about the questions mailing list