[ntp:questions] Re: Peer configuration question?

Nicholas Garfield Nicholas.Garfield at cern.ch
Mon Feb 2 14:43:28 UTC 2004

Having read the Mitigation rules I think the prefer keyword is OK to use whether peering or not.
"In order to provide robust backup sources, primary (stratum-1) servers are usually operated in a diversity configuration, in which the server operates with a number of remote servers in addition to one or more radio or modem clocks. In these configurations the suite of algorithms used in NTP to refine the data from each peer separately and to select and combine the data from a number of servers and clocks. As the result of these algorithms, a set of survivors are identified which can presumably provide the most reliable and accurate time. Ordinarily, the individual clock offsets of the survivors are combined on a weighted average basis to produce an offset used to control the system clock.
However, because of small but significant systematic time offsets between the survivors, it is in general not possible to achieve the lowest jitter and highest stability in these configurations. This happens because the selection algorithm tends to clockhop between survivors of substantially the same quality, but showing small systematic offsets between them. In addition, there are a number of configurations involving pulse-per-second (PPS) signals, modem backup services and other special cases, so that a set of mitigation rules becomes necessary to select a single peer from among the survivors. These rules are based on a set of special characteristics of the various remote servers and reference clock drivers specified in the configuration file."

The above explains why I started using a prefer statement.  Having observed the behaviour after activating the "peering" I can say with confidence that the peering has not affected the stability of either my stratum 1 or stratum 2 clocks. cool :-)

  "Richard B. Gilbert" <rgilbert88 at comcast.net> wrote in message news:efmdnXg5tqNyQ4fd4p2dnA at comcast.com...
  Well, I'm less certain of my ground here but here goes.

   I believe that stratum takes precedence over preference.  (someone please correct me gently if I'm wrong).  A server with a lower stratum will be used before a higher stratum that is preferred.

  Peers and servers within the same stratum start out as equals.  The selection algorithms will be applied equally to all of them.  Remember that the only difference between peers and servers is symmetry or asymmetry in their relationship with you.

  Read RFC1305 or as much of it as you can understand.  Read the documentation for the reference implementation, and observe your servers and clients in action.  When you see behavior you don't understand, ask.

  Nicholas Garfield wrote:


Thanks for such a concise reply.  You have summed up nicely what I wanted
confirming about peers.....
so how does the "prefer" statement effect this scenario?  Suppose you have a
preferred "server" and a non-preferred "peer".
Will the server always win?  Suppose neither are "preferred", will the
"peer" always take precedence over a server?
I realize this is probably just pedantic, but I really want to understand
the dynamics of such a system.



"Richard B. Gilbert" <rgilbert88 at comcast.net> wrote in message
news:OcednU09jIRVAITdRVn_iw at comcast.com...
  The difference between peers and servers is very simple!

Peers have a symmetric association; either one can supply time to the
other.  If two peers have different opinions as to what time it is, they
will try to determine who has the more accurate time and both will use
that time.

Servers are asymmetric; a server can supply time to you but will not
accept time from you.

For maximum reliability, you would have your two stratum 1 servers
configured as peers.  If everything is working properly, these two
servers should track each other and the "correct" time within a few
microseconds.  If either GPS receiver fails, the server it is connected
to will obtain the correct time from its peer and continue to serve its
clients.  If you have a network neighbor who also operates stratum 1
servers you might wish to peer with his servers as well.  A setup like
that gives you both a defense in depth against failures of reference
clocks, or servers.

Configure your stratum 2 servers to peer with each other and to be
served by both stratum 1 servers.  All clients should be served by the
stratum 2 servers.

Your single point of failure would be the GPS satellite system itself.
Highly unlikely but the DOD could just pull the plug on all those
satellites. . . .

Nicholas Garfield wrote:

    Hi All,

I have two stratum 1 servers which are each connected to Trimble Accutime
GPS receivers.  Then two more servers use these sources as stratum 2
I have been reading the NTP FAQ Section 6.2. Advanced Configuration,
  I am unsure of the configuration.

(1) For the stratum 1 servers, should there be a peering relationship
between them?

ie on server 1 config is

server prefer minpoll 4 maxpoll 4 #ACCUTIME
peer version 4 minpoll 4 maxpoll 4 #SERVER 2

On server 2 config is

server prefer minpoll 4 maxpoll 4 #ACCUTIME
peer version 4 minpoll 4 maxpoll 4 #SERVER 1

Also, would someone be kind enough to clarfiy whether or not the "prefer"
keyword is necessary here.  I figure (probably incorrectly) that I would
prefer the time to come from the accutime, but how would this work with
  idea of selecting the "best" clock.

(2) Similarly the config for the stratum 2 servers uses peering.

I am not 100% clear what the difference between a "server" declaration
      and a
  "peer" declaration is.  When a server and a peer is specified how does
  choose which is preferred?

Thankyou in advance to anyone who can explain clearly when/where the peer
and prefer statements should/should not be used.





More information about the questions mailing list