belinfan@cs.utwente.nl (Axel Belinfante) (12/04/90)
Dear ntp-knowledgeable,
I found your adress in the documentation of the ntp package that i ftp-ed
from trantor. At our (part of our) site, the Department of Computer Science
at the University of Twente in the Netherlands, we are thinking of using ntp
to synchronize the clocks of our workstations, and we would like to ask
you for some suggestions on the configuration.
To sketch the context: we have a little over one hundred workstations,
part of them disk-less, currently most sun 3/60 or sun4, a dozen hp s9000/300
workstations, and a few others (eg. decstation).
Those workstations are served by about a dozen file servers.
We would like to start with a small number of machines (3 to 5) that
use ntp, as a kind of test configuration, and later on run ntp on all the
machines of the department.
The documentation in the ntp package gave me the impression that
for the configuration of the entire department the best thing to do is
to set up a small `network' of machines that synchronize each other and
get their time info from the Internet, and let those `networked' machines
(re)distribute the time to the bulk of the machines.
To be honest, i got a little confused when i read the ntpd man page
when it, after introducing peer, server and passive time server
specifications, started talking about a `client' keyword.
The document `Internet Time Synchronization: the Network Time Protocol'
mentions a couple of modes, the ntp(d) man page mentions a couple of modes,
but i'm not so sure how they match, and which we of them we should use.
So, i would like to ask the following questions:
- sofar i have found the ntp(d) man pages, and two articles
about ntp (`Internet Time Synchronization: the Network Time Protocol'
and `Network Time Protocol (Version 2) Specification and Implementation',
both by D.L. Mills)
Does a document exist that deals specificly with configuration issues?
Section 6.2 of `Internet Time Synchronization: the Network Time Protocol'
deals with this rather briefly.
- how should we configure the testing configuration of 3 to 5
workstations? (preferably in such a way that it will be easy to
scale up to a configuration for the entire department.)
- how should we configure the entire department?
- to which ntp servers should we connect for the time info
(although they will be present in a configuration *file*, the
*names* of those ntp servers are not necessarily part of the
configuration: the configuration can simply say: specify
n remote ntp servers of stratus s with keyword/mode m)
- which version of ntp software should we use, where can we get it,
how can we get info about updates/patches, does a mailing list exist,
or will such information be posted to a USENET newsgroup?
(the version i have says in the README file:
README for UNIX NTP release $Date: 89/05/18 12:43:58 $ $Revision: 3.4.1.6 $
With kind regards, and thanks for your assistence,
Axel Belinfante (belinfan@cs.utwente.nl)
dept. of Computer Science, Univerity of Twente, the Netherlands.Mills@udel.edu (12/06/90)
Axel, Answers to many of your questions can be found by browsing the directory pub/ntp on louie.udel.edu. The two daemons ntp.3.4 and xntp distributions can be found there as well. There are three RFCs about the protocol, all in that directory and in PostScript format. All of the RFCs have been revised and either published in journals or distributed for additional review and comment. I have been promoting NTP Version 3, the specification of which is also in that directory, but so far only the LSI-11 fuzzball time servers speak that tongue. Others on this list may have more specific veiws and comments. You have primary time servers west, east and north of you in Europe. See the file clock.txt for details. Dave