[time-nuts] NIST UT1 NTP server results
Majdi S. Abbas
msa at latt.net
Fri Jul 22 18:39:12 EDT 2016
On Fri, Jul 22, 2016 at 06:23:07PM -0400, John Hawkinson wrote:
> With respect to interpolation and soforth, it seems like a lot of NTP
> cares more about frequency than offset, and all this stepping presumably
> wreaks havoc with the frequency? Maybe I'm wrong though...
While fudging the reference ID would be nice, I don't think
anyone is going to find this server without looking hard for it.
Even if they did, they should be running multiple upstream reference
clocks, and should filter out the shift pretty easily.
As for the frequency loop, even in well conditioned datacenter
environments there's a noticable diurnal shift in temperature, and the
resulting shift in clock frequency. Your average PC clock is so bad,
that I doubt running it against a UT1 server is terribly noticable.
The rise of alternative schedulers, SMP, and cheap commodity hardware
make for some jittery system clocks indeed. The daily shift in the
provided time is well in that noise level.
More information about the time-nuts