[comp.protocols.time.ntp] choosing peers

trost@REED.REED.EDU (Bill Trost) (04/26/91)

The manual for ntpd(8) says:

     NOTES
          Please choose your NTP peers carefully; send mail to
          ntp@TRANTOR.UMD.EDU for assitance.

Reed is gatewayed onto NorthWestNet (a part of NSFNET) via a gateway
to washington.edu at ogi.edu.  I've asked for assistance from OGI as
well, but they seem to be busy with finals or something silly like
that.	:-)

Here are my to ntpd configuration files for the moment (following some
of the various README's).  We're only using two local stratum-2 hosts
at the moment (I may add a third when another appropriate machine
arrives).

These machines are running the ntpd supplied with DECstation Ultrix,
by the way.

- romulus:/etc/ntp.conf
	trusting no
	#
	# Primaries
	peer	clepsydra.dec.com
	peer	fuzz.sdsc.edu
	peer	wwvb.erg.sri.com
	#
	# Secondaries
	peer	fuzz.sura.net
	peer	libra.rice.edu
	#
	# Local (secondaries)
	peer	vishnu.reed.edu


- vishnu:/etc/ntp.conf
	trusting no
	#
	# Primaries  
	peer	ncarfuzz.ucar.edu
	peer	norad.arc.nasa.gov
	peer	wwvb.isi.edu
	#
	# Secondaries
	peer	fuzz.psc.edu
	peer	lilben.tn.cornell.edu
	#
	# Local (secondaries)
	peer	romulus.reed.edu

Thanks for your help.

hakanson@CSE.OGI.EDU (Marion Hakanson) (04/26/91)

Bill,

>Reed is gatewayed onto NorthWestNet (a part of NSFNET) via a gateway
>to washington.edu at ogi.edu.  I've asked for assistance from OGI as
>well, but they seem to be busy with finals or something silly like
>that.	:-)

Sorry I gave you the brush off earlier, but you didn't say you were in
a hurry!  If you really can't wait....  We do have a (budding) NTP
hierarchy within NWNET which I coordinate in my spare time.  The
general model we follow is that each campus sets up one (or two, or
three, for larger sites) stratum-2 server, each of which has "server"
relationships with two stratum-1 clocks from clocks.txt, as well as
"peer" relationships with two or three other NWNET stratum-2 servers.
The network of NWNET stratum-2's is being (loosely) organized.
Currently the only official one is my own clocktower.ogi.edu, but
we're in the process of designating two or three at ORST (Oregon State
Univ.), and there's a couple at UWashington and a couple at UOregon.
This is why I asked you to wait in our private correspondence.

>Here are my to ntpd configuration files for the moment (following some
>of the various README's).  We're only using two local stratum-2 hosts
>at the moment (I may add a third when another appropriate machine
>arrives).
>
>These machines are running the ntpd supplied with DECstation Ultrix,
>by the way.

Three s-2's is overkill for your campus.  We only run one here,
although I'm considering a second as backup.  We also set up a couple
s-3's in each department, "peer" those with our campus s-2 and with a
couple NWNET s-2's, giving us a plenty robust hierarchy.  All
departmental workstations are set up as NTP "clients" at s-4, in that
they peer at their respective departmental s-3(s) and with no others.
Keep in mind that we don't want to beat up on the poor s-1's, OK?

Also, I recommend that your s-2 server run xntpd instead of ntpd, if
at all possible (since the former gives you more info if you want to
monitor it), but ntpd is adequate just the same.  The Ultrix-4.1 ntpd
on our DEC's seems to behave OK, but ours are only "clients".

So, I recommend that you set up just one s-2 server at Reed, point it
at two stratum-1's (don't use ncarfuzz -- everyone here but ORST has
an asymmetric route to that ticker), and at a couple other NWNET
s-2's.  You're welcome to use our clocktower as one of them, and then
probably also check with UW or ORST for another one or two.  Having
more s-2's doesn't really help you out much, since you only have one
network connection, and your departmental s-3's protect you against
your s-2 going insane.  I think there are enough NWNET s-2's that you
don't need to go "outside" to peer with an s-2 server.

Anyway, I keep hoping to hear that someone inside NWNET has bought
themselves a good radio clock (not a Heath job), so we could have
us our very own stratum-1 in the NorthWet, but so far no soap....

-- 
Marion Hakanson         Domain: hakanson@cse.ogi.edu
                        UUCP  : ogicse!hakanson