[ntp:questions] delay in stats

Danny Mayer mayer at ntp.isc.org
Fri Mar 2 14:02:40 UTC 2007


edouard funke wrote:
> Hi,
> 
> I am running some tests on the NTP protocol for my company : I need to
> achieve (if possible) in LAN a precision (for clients) of about 1
> millisecond (or less :-))

To get that kind of precision you need to have the system directly
connected to a refclock. You'll rarely get that kind of precision over a
network.

> 
> 1. the following properties are studied :
> 
> - the time for a client to "calibrate" its clock
> o with/without iburst mode
> o with/without drift file
> o with fixed drift

There is no such think.

> o with different values for max polling
> 

It will calibrate without ever reaching maxpoll.

> - the maximum precision achievable for clients :
> o with different values for max polling

You will get surprising results if you change maxpoll. The algorithms
are designed to give you the best results without touching maxpoll. You
really need to understand the engineering and algorithms used before you
can draw any conclusions on the best value for maxpoll.

> o with huff n'puff filter

Are you expecting asymmetric paths?

> o with all clients in « peer » mode or at least some of them
> 
> the NTP  server clock reference will be :
> o undisciplined local clock

Then forget precision here.

> o internet stratum 1 or 2
> o GPS (i have a meinberg lantime M300 and PCI card + GPS receiver on loan)
> 
> I am trying to know here which configuration parameters can improve
> the "calibration" speed and maximum precision of clients clock.
> I am not trying to have the exact time on clients but rather, that all
> clients have the same time.
> 
You are better of reading Dave's research reports on how all of this
works and his analyses of the data.

> I must run NTP daemon on Windows Server 2003 SP1 and it almost sure
> that, in the end, NTP servers wont be able to have internet connection
> or even GPS : they are included in tests because i want to determine
> the impact of the server precision on the client precision.
> 

Well in such a case your server has no precision since it's not
connected to anything and is just drifting.

> For the GPS, i know there is some hardware solution that can run in
> free-run mode after 24h calibration (meinberg).
> 
> I am sure it is a commun problem but i couldn't find any good reports
> on it : if you have some that might interest me, that would be great
> :-))
> 

Did you try http://www.eecis.udel.edu/~mills/ntp.html

> 2. I just checked the ntpd 4.2.4 changelog : is the windows multimedia
> timer enabled by default ? Where can i find some informations on it ?
> 

You need to add -M to the command line or the service to enable it. It's
not enabled by default.

Danny



More information about the questions mailing list