[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: ntp setup for a local server and a few clients to that server

From: "Keith Morse" <kgmorse mpcu com>

> A couple o' nits that I've developed whilst watching messages about
> setting up ntp. 
> I would hesitate recommending level 1 servers.  Personally I try to find
> level 2 or 3 servers for service.  Also you might ask your ISP if you can
> point to their time server, assuming they maintain one.  Geographically,
> (or rather network topoligically) close is bestest.  Traceroute can help
> you there.

At one time I added the gtei.net routers through which I connect. I discovered
they run NTP. I also discovered they all have a systematic error from most of
the other sources I use. I discontinued using them. You DO have to weigh
these things, ya know. (Of course this could be the gtei routers and name
servers (not all "local") having more "symmetrical routes" to and from
thus making their times more accurate. The sign of the error observed prompted
me to question that hypothesis. {^_-}) The other observations are right on.

> Three servers is what I remember being recommended.  Somewhat GPS, the
> more sources you have, the more accurate the time and redundancy will be
> assured.

Three seems to be sufficient. When asymmetrical routing happens the time
becomes more confused rather than more accurate. For REAL precision you need
to analyze further what you are seeing. (Heh, when I tried with the Dirtlink
servers by hunting around for "things" running NTP on the Pasadena Dirtlink
campus addresses I discovered they, too, were serving quite different time
from what I got from UCSD.)

Most people do not need to worry about THAT level of accuracy. I "wondered"
and explored it a little. I settled on something easy and seemingly fairly
accurate. Low tens of mS accuracy is sufficient for any of my needs here.
I was a LOT more concerned when I designed some of the GPS satellite hardware
that is in the path between the precision oscillators and the entire rest of
the satellite, though. The "time improvement" you see with a lot of similar
quality (route and "level") is in no way "GPS" like. But then, that's not a
discussion for this place and time. So I won't tell my tale about a certain
Air Force captain, the project manager, me, and a couple other engineers
about "dithering" and "differential GPS" and an intense desire NOT to hear.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]