On Mar 26, 2009, at 14:52, Scott Baker wrote:

> Is everyone else seeing much LOWER offsets on the graphs since the  
> move? My
> offset used to hover between -10 and 40, and not it hovers between  
> -6 and -2.

Adrian correctly guessed half of the solution:

1) The NTP Pool server was sharing a 10Mbit connection before with  
perl.org.  Occasionally that'd get busy enough to create a bit of  
latency/jitter.  Over the last couple of weeks the pool servers and  
perl.org each have a 100Mbit link (thanks to http://www.phyber.com/ )

2) The check script is really a simple SNTP client -- it checks the  
time against a "known good" server.  Before that was just one of my  
ntpd's on a regular Linux or FreeBSD box (I forget which) -- now it's  
an ntpd on an awesome Meinberg lantime server with a GPS antenna on  
the roof.  (Which obviously gives a much better "baseline" to check  
against).


   - ask
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers

Reply via email to