[ntp:questions] Re: Clock selection?

David L. Mills mills at udel.edu
Sun Jan 18 16:37:39 UTC 2004


Richard,

See the clustering algorithm slide in the architecture briefing at the
NTP project page linked from www.ntp.org. See also the error budget
slide in the same briefing. The concepts are really simple: order the
servers by stratum and synchronization distance and toss out the
outlyers until only three survivors remain. The devil of course is in
the details.

Dave

"Richard B. Gilbert" wrote:
> 
> Can anyone explain the clock selection here?  Why would the clock with
> the highest delay, offset, and jitter be selected?  To my unpracticed
> eye, 128.59.39.48 appears to be the cleanest source of time!
> 
> sunblok_# ntpq -p
>      remote           refid      st t when poll reach   delay   offset
> jitter
> ==============================================================================
> *louie.udel.edu  128.4.40.10      2 u  682 1024  377   45.499    9.590
> 11.502
> +filbert.cc.colu 128.59.39.48     2 u  645 1024  377   14.715    2.824
> 0.757
> +otc2.psu.edu    128.118.25.12    2 u  806 1024  377   22.974   -4.068
> 1.317



More information about the questions mailing list