[ntp:questions] NTP stratum-1 architecture

Harlan Stenn stenn at ntp.org
Wed Dec 5 19:45:47 UTC 2007


>>> In article <0ED191612CF19246BF2E48B12CB0C719167755 at nl-ex006.groupinfra.com>, raymond.groenewoud at logicacmg.com (Groenewoud, Raymond) writes:

Raymond> Folks, For a large organisation, I am reviewing the current
Raymond> NTP-infrastructure. Because high-availability is a major concern, I
Raymond> am considering the following architecture (improvement):

I recommend http://support.ntp.org/Support/DesigningYourNTPNetwork if you
have not already seen it.

There are tradeoffs.

And you were saying that GPS would cost you more than rubidium?

The problem with only 3 S1s is that if 1 is unreachable and if the other two
drift apart for any reason, there is no way to decide which of the other 2
to believe.  If you get 4 S1s and for whatever reason they form 2 two-unit
cliques, your downstreams will have no way to know which clique to believe.

These problems may not be all that bad, however.

You may be able to add a modem refclock during times of "trouble".

And I would also recommend adding some offsite NTP servers to your mix, just
to keep your server numbers up.

If you have good monitoring software in place then if there is ever a
problem people can be notified and steps taken before things get out of
hand.

It all boils down to how many 9s you want and can afford.

H
-- 
http://ntpforum.isc.org  - be a member!




More information about the questions mailing list