[ntp:hackers] Precise synchronisation on a local network

P.O. Gaillard pierre-olivier.gaillard at fr.thalesgroup.com
Mon Jul 3 15:59:12 UTC 2006


I am currently using NTP servers using a GPS clock to synchronise PCs on local 
networks (fedora Core 3, ntp-4.2.0.a.20040617-4).

This setup is quite different from the regular NTP setup since :
  - we have few clients (2 or 4 PCs querying 2 servers)
  - the Round Trip Time between client and NTP server is very low (600 to 700 
  - we want a precise synchronisation (better than 2ms)
  - we buy lots of PCs and cannot rely on their clock to be high quality (some 
have 10ppm clocks, other 100ppm ones)

To improve the accuracy of the NTP client we have  :
  - set minpoll and maxpoll to 4 (16s) to accelerate convergence and increase 
the number of measurement performed
  - used iburst to accelerate the convergence of the NTP client at startup

Unfortunately we observe that on PCs with 100ppm clocks NTP does not keep time 
very well. The error can reach up to 10ms from time to time. This seems very 
disappointing since the measurements have a precision better than 350us (700/2).

Is there a standard and simple way to tell NTP to trust measurement a little 
more and avoid letting the clock drift that far ?

   thanks for your attention,

   P.O. Gaillard

More information about the hackers mailing list