[ntp:questions] Tighter regulation?
unruh at invalid.ca
Fri May 24 14:55:13 UTC 2013
On 2013-05-24, Mischanko, Edward T <Edward.Mischanko at arcelormittal.com> wrote:
>> > On 2013-05-21, Mischanko, Edward T <Edward.Mischanko at arcelormittal.com>
>> >> My concern is that too much data is being thrown away when polling
>> >> above 256 seconds and that allows excessive wandering of my clock.
>> If too much data is being thrown away, it would be because the poll
>> adjust algorithm has chosen too high a poll interval (or you have too
>> high a minpoll) for the noise statistics.
> [Mischanko, Edward T]
> This is exactly what I am saying!! This needs to be corrected; I consider it a bug.
And until you take over the running of ntp, it will remain your opinion,
and not that of the supporter.
> The samples are, effectively,
>> put through an IIR digital low pass filter, with a bandwidth that is
>> sufficiently narrow that variations across the 8 samples are largely
>> suppressed. However, if you include all the samples, their overall
>> average is likely to be biased away from a zero error, and as that is
>> likely to not change, or only change slowly from one group of 8 to the
>> next, it will introduce an error that is not filtered out.
>> It is not throwing out anything like the equivalent of 7 out of eight
>> samples, because the contributions of the variations between samples are
>> high frequency noise, which is probably being attenuated by a factor
>> significantly more than 8.
>> The way that ntpd chooses the poll interval is by looking at the offset
>> and jitter (exponential average of sample to sample variations). If the
>> offset consistently exceeds the jitter, it means that part of the offset
>> cannot be accounted for by the measurement noise, and more aggressive
>> correction is needed. Unfortunately, if something make the crystal
>> frequency change suddenly, after a long stable period, it can take some
>> time before it realizes it needs to do aggressive correction.
> [Mischanko, Edward T]
> It takes too long to figure out it needs a more aggressive correction.
> If I leave maxpoll at the default of 1024 seconds, my clock drifts outside
> of 5 milliseconds consistently.
So lower maxpoll!
> Too much assumption is made that everyone will have the perfect computer and
> the perfect network when configuring these various filters. What works on
> the blackboard does not always work in reality. My computer has a
> -19 precision but it can't keep time inside 1 millisecond with default
> Settings; go figure.
That is unfair. Mills has spent a lot of time making sure it works in
the real world. That his priorities are not yours is surely something he
is entitled to. Why do you not use chrony? I know you have tried it.
>> questions mailing list
>> questions at lists.ntp.org
More information about the questions