[ntp:questions] Re: why difference about 200ms with time.nist.gov

Richard B. Gilbert rgilbert88 at comcast.net
Wed Mar 29 12:53:49 UTC 2006


flyfatpig at gmail.com wrote:

> I built a time server with redhat 7.2 and ntp 4.2.0 and Motorola M12T.
> The time server works well.But when I  sync with time.nist.gov ,I found
> my time server is about 200ms later .who can tell me why?
> 
> 
> [root at gt4 lxh]# ntpdate -q 192.168.2.7
> server 192.168.2.7, stratum 1, offset 78.759011, delay 0.02599
> 29 Mar 15:37:32 ntpdate[1090]: step time server 192.168.2.7 offset
> 78.759011 sec
> 
> [root at gt4 lxh]# ntpdate -q time.nist.gov
> server 192.43.244.18, stratum 1, offset 78.958525, delay 0.26364
> 29 Mar 15:38:24 ntpdate[1092]: step time server 192.43.244.18 offset
> 78.958525 sec
> 
> [root at gt4 lxh]# ntpdate -q clock.isc.org
> server 204.152.184.72, stratum 1, offset 78.980319, delay 0.20029
> 29 Mar 15:45:16 ntpdate[1111]: step time server 204.152.184.72 offset
> 78.980319 sec
> 
> where 192.168.2.7 is my time server .
> 

time.nist.gov is generally a very difficult server to get good time 
from.  Everybody believes, because it is stratum 1 and operated by NIST 
that it must be a good server.

In a sense, they are right; that atomic clock is correct by definition! 
  It is the legal standard for time in the US.

OTOH, the server, and its network connection, are usually overloaded to 
the breaking point or beyond.

Next, you have in no way demonstrated that you (or NIST) are off by 200 
milliseconds!  You are showing an offset of 78 milliseconds.

Configure your server to use five or six internet servers in addition to 
  your Motorola Oncore.  Let it run for a day or two and plot offset vs. 
time of day for your Oncore and each network server.

Take an ntpq "banner"  (ntpq -p).   This will show you delay and jitter 
figures for the servers you have configured.  Low numbers for delay and 
jitter are good; high numbers are bad.

I graph my offsets for each server and I see that each server has a 
systematic offset from GPS ranging from two to five milliseconds.  I 
also see the effects of network congestion; during the daylight hours 
the internet servers tend to perform very poorly; they settle down and 
look much better during the hours of darkness.

This will give you a very good idea of which are good servers and which 
are not.  This will be specific to your site.  You may need to try quite 
  few; like the lady said, you have to kiss a lot of frogs to find a prince!




More information about the questions mailing list