[ntp:questions] ntpq +rv
David L. Mills
mills at udel.edu
Wed May 30 17:35:19 UTC 2007
Danny, This question has come up several times in recent weeks. I don't
know why it was a secret for so long.
What is (unfortunately) called noise in the rv billboard is the
exponential average of the first-order clock discipline differences.
This is the control mechanism used to adjust the poll interval and
perhaps the best measure of the peformance as a whole. Put another way,
if the clock discipline offset is the maximum likelihood estimate of the
clock error, the noise is the expected error in this estimate.
What is (unfortunately) called error in the billboard is the result of
the intricate error budget described in the book, NTPv4 reference and
implementation guide on the project page and the current NTPv4
specification draft. This statistic is a compilation of errors due to
very many causes and is intended as a thermometer to assess the
performance of the various algorithms. Think of error as the raw
thunderstorm battering the client and noise as the constant rain after
the mitigation algorithms.
Please tell your friends.
Danny Mayer wrote:
> Laurent Archambault wrote:
>>With the command ligne "ntpq -c rv", it's possible to see the variable
>>not for localhost and not for old version of ntpq.
>>But i don't known what is this "noise" for the NTP, and i must explain this
>>for make an private network.
> That's an excellent question. Initially I thought you meant jitter but
> looking at the -c rv output those are different. I don't see it
> documented anywhere.
> Looking at the source code it seems that noise is defined to be the
> clock offset jitter while jitter seems to be the system jitter. I can't
> explain this any better, I can only look at the source code. Dave should
> be able to provide a good explanation.
>>Note : the documentation on "ee.udel.edu" don't explain this for "ntpq"...
>>Thank for all.
> questions mailing list
> questions at lists.ntp.isc.org
More information about the questions