Date: Sun, 16 Jun 1996 10:55:10 +0200 From: Mats Lofkvist <mal@algonet.se> To: nate@sri.MT.net Cc: current@FreeBSD.org Subject: Re: NTP gurus Message-ID: <199606160855.KAA27613@aristotle.algonet.se> In-Reply-To: <199606150254.UAA23487@rocky.sri.MT.net> (message from Nate Williams on Fri, 14 Jun 1996 20:54:58 -0600)
next in thread | previous in thread | raw e-mail | index | archive | help
[...]
Understanding that my server may not be completely accurate, I am still
using it to keep the rest of the machines in my domain in sync. At most
they are a half a minute off from each other, vs. 10's of minutes in the
past.
Eh, do you mean milliseconds, not minutes? I would consider half a _second_
very bad with ntp running. A few milliseconds is good.
Anyway, back to my question. How do I determine 'how accurage' my
system's clock is based on /etc/ntp.drift? What are good numbers to
have?
[...]
The drift value is the amount ntp is skewing the system clock frequency
in parts per million. I got the following results when tweaking
TIMER_FREQ on my 486 (from my config):
# value ntp drift
# 1193182 -550 ppm?
# 1193634 -180 ppm
# 1193849 +- 3 ppm
#
options "TIMER_FREQ=1193849" #Improve the clock accuracy on Garm
I don't think the drift value matters very much when ntp is running,
as long as ntp is able to lock in the first place (my ntp couldn't lock
with the default TIMER_FREQ, so I had to fix it).
(When _not_ running ntp, it is of course nice to have the system clock
as accurate as possible.)
_
Mats Lofkvist
mal@algonet.se
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199606160855.KAA27613>