[ntp:questions] Number of Stratum 1 & Stratum 2 Peers

Rob nomail at example.com
Tue Dec 2 09:00:52 UTC 2014

edstuart at gmail.com <edstuart at gmail.com> wrote:
> Hi,
> I am looking to implement an NTP network and I was reading http://support.ntp.org/bin/view/Support/DesigningYourNTPNetwork.  It suggests 4 stratum 1 peers and 4 stratum 2 peers well.  I understand, that the document recommends this so that if one of the servers in a stratum dies, there will still be 3 servers from which a majority clique should be found.  The problem is that we will be starting out with simply 100 nodes at most.  
> At the same time, we want drift less than 1 second.  However, over the next 7 years or so, we should hit ~ 1000 NTP clients.  Is the document's recommendation overkill for my situation?

The number of servers is (at that scale) not related to the number of
clients, i.e. one server can easily serve 1000 clients.

If you want more than one server, and how many you want, only depends
on the reliability of your server(s) and the reliability you want to
get on your clients.  A single server works well to keep clients within
1 second, but when it fails of course the clients will start to drift
while you are fixing it.  The rate at which they drift depends on their
operating system, the quality of the hardware, and the stability of
the ambient temperature.

So depending on your requirements (how hard that 1 second limit is) and
the quality of your network management (how quickly can you bring a
failed server back up) you may need 2 or more servers to guarantee your

The whole "have 3 servers to select a majority" thing is absolutely not
required when your servers are accurately synchronized themselves and
your requirements are "only" within-a-second.  It is true that when you
have two servers the clients cannot know which one is right, but it is
trivial to keep servers within a millisecond of eachother with GPS and
within 10 milliseconds using only network peering.  To that is two
orders of magnitude better than you require.

More information about the questions mailing list