[ntp:questions] Synchronize clocks without a reference clock

Per Hedeland per at hedeland.org
Mon Mar 26 00:59:51 UTC 2007

In article <46054BE2.5070305 at ntp.isc.org> mayer at ntp.isc.org (Danny
Mayer) writes:
>Per Hedeland wrote:
>> In article <slrnf04vg8.4ed.kostecke at stasis.kostecke.net> Steve Kostecke
>> <kostecke at ntp.isc.org> writes:
>>> Here's the configuration you need:
>>>> [1.] participants are identically configured both as broadcast client
>>>> and as broadcast server.
>> So the part that says "operates in subnet configurations in all modes"
>> is wrong? Do you know this for a fact and/or have tried it? I would
>> imagine at least that if all the participating servers have each other
>> configured as "peer", there shouldn't "technically" be a need for a
>> broadcast setup.
>The biggest problem with peer is that you have to specify each and every
>system you want to have participate in the scheme. With broadcast and
>multicast you don't need to do that, they find each other.

Yes of course, and I guess this is appealing in some scenarios - in
others it may be largely irrelevant, e.g. I'd think a typical usage case
would be for someone setting up 2-4 "main internal" servers that get
their time from Internet servers and and provide it to lots of internal
clients. If you configure all those clients with the identity of the
"main" servers anyway, dropping it into their own config files too is
not a big deal, and it may be preferrable to setting up

Actually you'd likely want to configure those "main" servers as peers to
each other even if you weren't using orphan mode, thus orphan mode just
becomes a simple addition for increased reliability in case your
Internet link goes down - and much better than the "local clock with
minimum of 2 stratum difference" hack that has been the standard
recommendation, of course.

--Per Hedeland
per at hedeland.org

More information about the questions mailing list