[c-nsp] ntp

[Mack McBride]

As Gert said, it does vary what is 'reasonable'.

Reasonably accurate for a stratum 2 is going to depend more on latency than the server.
Particularly if the server is any appreciable distance away from its stratum 1 sources.
Within 1 millisecond is usually pretty good.
More than 5 milliseconds usually indicates a crappy network connection to the sources (too far or too unstable - high jitter).
If the clock on your server is very inaccurate then you could exceed those figures.

If you want more accurate than that you really need your own stratum 1s.

A CDMA based device is <10 microseconds to UTC when locked (<10 microseconds is required for CDMA cell phones to work).
A GPS based device is <30 nanoseconds to GPS when locked (GPS itself is <100 nanoseconds to UTC).

The drift rate of the clock is also important for stratum 1s.

A regular clock is going to be 10milliseconds / day on signal loss.
An OCXO is going to be 100 microseconds / day on signal loss.
A rubidium (similar to what is used in a cell tower) is going to be 8 microseconds / day on signal loss.

Mack

-----Original Message-----
From: cisco-nsp-bounces at puck.nether.net [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Gert Doering
Sent: Monday, June 25, 2012 12:55 PM
To: Nick Hilliard
Cc: cisco-nsp at puck.nether.net; chris stand
Subject: Re: [c-nsp] ntp

Hi,

On Sun, Jun 24, 2012 at 08:34:06PM +0100, Nick Hilliard wrote:
> On 24/06/2012 18:08, chris stand wrote:
> > Let them get time from your edge device(s) and let the windows 
> > clients get it from them - which I think they are already doing.
> 
> I got the impression that the OP had ~10k network devices which 
> required reasonably accurate time.  I.e. not client boxes.

"reasonably accurate" means different things to different people...

To an NTP geek, or someone who has to clock a synchronous network, it's somewhere in the 10^(-6) or better range.

To someone who just wants logged timestamps with the correct day+month in, everything that's not off by more than a few seconds is good enough... :-)

gert


--
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert at greenie.muc.de
fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de