I would not recommend using Cisco devices as NTP servers if the time is 
important to you.  There are bugs in the code that result in the offsets not 
getting any better than a few ms.  Using the prefer command works around one of 
the problems but it still wanders around a lot.

I point all my network equipment to our own dedicated NTP servers which can 
handle thousands of requests/sec which is a lot more that would ever be 
generated.

Cisco has a new implementation in the 12.4T train that is a big improvement.  
This implementation finally lets you set the min and max polling rates which is 
making a big difference.  I can get sub-ms offsets. There is still an issue 
where the accuracy is better with a single server than with two:

r01(config)#ntp server 10.1.1.1 ?
  burst    Send a burst when peer is reachable
  iburst   Send a burst when peer is unreachable
  key      Configure peer authentication key
  maxpoll  Maximum poll interval
  minpoll  Minimum poll interval
  prefer   Prefer this peer when possible
  source   Interface for source address
  version  Configure NTP version
  <cr>

David 

-----Original Message-----
From: [email protected] 
[mailto:[email protected]] On Behalf 
Of Towli
Sent: Monday, April 13, 2009 7:02 AM
To: [email protected]
Subject: [ntp:questions] Best Ntp Practice

Hello all

We have found a decent location for our 2 DCF77 boxe's and I am going to 
configure our environment for using these.
Would best practice be to configure 2 core (Cisco) switches to use the two 
DCF boxes like this: ntp server (ip of DCF 1) prefer, ntp server (ip of DCF 
2) and then hereafter i am considering entering a third "regular" reachable 
Ntp-server. Will the servers be prefered in the order they are entered on 
the switches?
I will then point all our other networking equipment to the 2 mentioned core 
switches for NTP data (in case you in here "ok's" it). How can i make sure 
that the core switches is handing out ntp data (is there a special command 
for this ?). The reason i am considering a setup like this is, that the DCF 
looses their stratum 1 status when (if) they loose connectivity to 
Frankfurt. With the setup i describe only the 2 cores switches needs to sync 
to another NTP-server (like DCF nr 2 or the 3.rd configured server)
I am realy looking forward to a comment as i am a newbie on this.

Best regards from Denmark
/Towli 

_______________________________________________
questions mailing list
[email protected]
https://lists.ntp.org/mailman/listinfo/questions
_______________________________________________
questions mailing list
[email protected]
https://lists.ntp.org/mailman/listinfo/questions

Reply via email to