NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/protocol/time/ntp/1100

Internet Message Format | 1992-12-14 | 2.7 KB

Path: sparky!uunet!zaphod.mps.ohio-state.edu!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!spool.mu.edu!agate!ucbvax!UDEL.EDU!Mills
From: [e-mail]
Newsgroups: comp.protocols.time.ntp
Subject: Re:  assistance for selecting a peer..
Message-ID: <[e-mail]>
Date: 14 Dec 92 05:08:52 GMT
Sender: [e-mail]
Distribution: inet
Organization: The Internet
Lines: 44

Sanjay,

First, let me suggest you use xntp3 instead of ntpd on your principal
server, as this version is much more robust in the face of likely
congestion on your access paths to extra-national primary servers.
A couple of years back I was the guest of IIT Bombay on behalf the
UN Development Program, so I have some idea of what you are up
against. While I don't know the present access path, presumably
via your ERNET headquarters in Bombay, I suggest you run the
traceroute program to find out the particular route and inspect the
file clock.txt in the pub/ntp/doc directory fia anonymous ftp on
host louie.udel.edu for likely candidates. The complete distribution
of ntpd, xntp3 and related programs reside in the pub/ntp directory
and subdirectories on that host.

Please understand the robustness of time distribution depends critically
on redundancy and diversity, which means you need redundant servers
and diverse access paths to multiple servers "closer" to the primary
source of synchronization. The last time I visited India you had only
one access path. Even in that case you should consider multiple
NTP servers for your country. In view of the low network overhead of
xntp3, you might consider each of your four sites as redundant and
have them each peer with outside reference sources and with each
other. See the various information files in the pub/ntp/doc directory
and in the xntp3 distribution.

There is documentation on the tickadj program in the xntp3 distribution.
You need to specify a tick paramter of 9999 for those Sun platforms
known to me. There are example configuration files in the distribution
as well.

In your part of the world, the only precision time signals suitable
for connection to a computer are from the Global Position Service
satellites, which require a most expensive receiver, and the OMEGA
radioonavigation system (receiver made by TrueTime - see clock.txt),
which is only slightly less outrageous. There is a driver for the
OMEGA receiver in the xntp3 distribution (same as the GOES receiver).
I challenge an enterprising grad student to cobble up a receiver
for the BBC on-hour pips, which are I am told derived from precision
sources and readily receivable in your country. With propagation
corrections and selection of only the domestic UK transmitters, you
could realize a quite accurate national time source.

Dave