[ntp:questions] noise, frequency, stability, tai, rootdispersion, ...

Richard B. Gilbert rgilbert88 at comcast.net
Tue Feb 3 16:13:53 UTC 2009


alkopedia at googlemail.com wrote:
> Hi
> Where can I find a detailed description of all these values provided
> by "ntpq -c rv"?
> What do you think which values are important for logging and graphical
> display to see problems fast? Here is an example
> http://www.wraith.sf.ca.us/ntp/rrd/index.html but I'm not sure what is
> important because I can't find good documentation about all the
> values.
> 
> Greetings
> Thorsten

"Das Buch" (Professor Mills' book on NTP"), or comments in the source 
code for ntpd.  RFC-1305 or its successor if ever released.  Here's a 
link to "Das Buch" 
http://www.amazon.com/Computer-Network-Time-Synchronization-Protocol/dp/0849358051/ref=sr_1_1?ie=UTF8&s=books&qid=1233676373&sr=8-1

I have not seen it but I understand that the math can get pretty esoteric!

Suggestions:
Reach - the reachability of the server.  It's an eight bit shift 
register; a one bit is shifted in from the right each time the server 
replies to a query.  A zero is shifted in each time there is no 
response.  It's displayed in octal notation; 377 is good, any thing else 
means that the server failed to respond to one or more of the last eight 
queries.  Note that this is useless immediately after startup; this and 
  most other statistics are of little value until ntpd has been running 
for about thirty minutes!

Dispersion - the estimated difference between your clock and the clock 
at the root of the tree.

You will probably find that the statistics look their best between 2300 
and 0700 local time; people are sleeping rather than beating on the 
servers and this generally means that the round trip time for query and 
reply is short and more or less constant; ntpd really likes this and 
does its best work during these hours.




More information about the questions mailing list