[ntp:questions] Yet Another Isolated Network Question
brad at stop.mail-abuse.org
Thu Dec 16 22:23:29 UTC 2004
At 4:59 PM +0000 2004-12-16, Bruce Baumgart wrote:
> 1) It will remain completely, totally private. No outside connections,
> no matter how brief. Never.
> 2) I cannot simply add a GPS/WWVB/etc reciever. End of discussion.
Hmm. Sounds like the application might be for a classified
network, or at least one handling highly sensitive data.
In this case, were it me, I would be looking for a way to add an
accurate timing reference, such as an oven-stabilized crystal
oscillator (OXCO), or perhaps even a rubidium clock. There's no
connection to the outside world, so the absolute time is whatever you
set on the server according to your watch (or whatever). But at
least the drift will be relatively tightly controlled.
> 3) I cannot alter connection methods, topology, etc.
What does this mean? Does this mean you can't change what
software runs on what server? Does this mean that you can't
physically change what is plugged into what, even down to the level
of a mouse or keyboard, or a serial cable to an external timing
reference device which is not otherwise connected to anything else in
the world (other than a power source)?
> 4) I am not interested in sub-second synchronization. When I got brought
> in, the DNS server and the SMTP server were 1/2 an hour apart, and had
> been that way for a long time. +/- a few seconds is fine <shudder>.
I understand, but the situation that you inherited was caused by
drift. I would think that drift is something you'd care about, and
want to try to keep under reasonably tight control. Ideally, I would
think that you would want an external reference device to which ntpd
can compare against the internal local clock, as opposed to just
depending on the local clock.
> Will NTP work the way I think it will in these circumstances? Mainly
> thinking about the local server serving time, not drifting too far, etc?
I confess that my knowledge of NTP is not nearly as deep as you
will find in some other people on this group, but I would expect that
this should, more or less, work basically the way you want.
> Is there a better way, given my loose requirements and draconian
The one thing I'd look at would be adding an OXCO to the
server(s) as a reasonably accurate relative time reference -- they
won't pull data out of the airwaves (or any other method) to tell you
what the time currently is, but they'll give you a pretty good idea
of just exactly how long a second really is, and should help you keep
your drift to a minimum.
But maybe that's just me.
Brad Knowles, <brad at stop.mail-abuse.org>
"Those who would give up essential Liberty, to purchase a little
temporary Safety, deserve neither Liberty nor Safety."
-- Benjamin Franklin (1706-1790), reply of the Pennsylvania
Assembly to the Governor, November 11, 1755
SAGE member since 1995. See <http://www.sage.org/> for more info.
More information about the questions