Hushpuppy wrote: > I have a machine that is syncing to a Stratum-2 server which is > syncing to a Stratum-1 server, which has a GPS Stratum-0 device > attached to it. At my work (financial services) the market rules > mandate synchronizing our clocks to within a maximum 3 second > difference of the NIST atomic clock. We also need to log/document > that we are within 3 seconds.
If ntpd is allowed to step and it has been more or less correct at some time, you will not get such a large error unless something is broken. If something is broken, it will invalidate the statistical measurements, anyway. However, I think the worst case is something like root delay / 2 + root dispersion, plus last hop delay / 2, plus precision, + 15 microseconds for every second since the time was updated. You may need to use reference time for the last one. This may not be quite right but it includes most of the terms. The 15ppm is the figure that ntpd uses to calculate dispersion. I'm not 100% sure which side of the link is responsible for the servers contribution to root dispsersion. Something looks funny with your dispersion figures for the server; they don't seem to be reflecting the fact that its reference time is rather old. Note that "offset" doesn't mean error. Ideally offset should be about an order of magnitude larger than the actual error, although that is often not true. A couple of other observations. It is advisable to use at least four real servers and you should drop the local clock unless both this machine is serving time to other machines and you fully understand what the local clock driver does. _______________________________________________ questions mailing list [email protected] https://lists.ntp.org/mailman/listinfo/questions
