[ntp:questions] Re: New to NTPD

Richard B. Gilbert rgilbert88 at comcast.net
Wed May 10 00:29:15 UTC 2006


Ted Gervais wrote:

> 
> 
> Thanks Brian  for your input on this matter.  Sure need something right 
> now and your input surely looks helpful.
> 
> Now - what I have in place right now shows some improvement but I wonder 
> if it is right yet.  Here is what ntpq -p gives me:
> --------------
> Every 1.0s: ntpq -p                                     Tue May  9 
> 19:42:06 2006
>      remote           refid      st t when poll reach   delay   offset 
> jitter
> ============================================================================== 
> 
> +clock.via.net   .GPS.            1 u  223  256  377  101.718   21.183 
> 5.465
>  gnomon.cc.colum .USNO.           1 u  483 1024  301   55.229   26.100 
> 2.439
> -NAVOBS1.MIT.EDU .PSC.            1 u   11  256  377   60.820   28.740 
> 2.602
> +ntp3.usv.ro     .PPS.            1 u   15  256  377  173.654   26.977 
> 1.469
> *clock.xmission. .GPS.            1 u  216  256  377   88.717   23.674 
> 4.681
> +c-24-130-58-99. .GPS.            1 u   83  256  377  118.072   23.165 
> 1.358
> 
> -------------------------
> 
> How does this look?  I think it actually is working now?? Anyone, please
> let me know.
> 
> And Brian... I will try the list of servers that you provided. Thanks 
> very much for that input..
> 
<big snip>

You seem to have very large round trip delays to the servers you have 
chosen.  Shorter delays are better.  The error in transmitting time from 
server to client is bounded by one half the round trip delay.  With a 
delay of 173.654 milliseconds, all you know is that the error can't be 
worse than 86.3 milliseconds.

Look for servers close to you in net-space; i.e. those with low values 
for delay.  I look for delays of fewer than twenty milliseconds and have 
found five servers that meet this test.  If I could not find sufficient 
servers meeting the twenty millisecond objective, I would relax my 
standards until I could find five servers.

ntpd has selected a server for synchronization, clock.xmision.  The 
asterisk preceding the server name indicates that it has been selected. 
  The offsets are still large but I don't think you had been running 
very long when you got this output.




More information about the questions mailing list