[ntp:questions] Synchronize clocks without a reference clock
Richard B. gilbert
rgilbert88 at comcast.net
Mon Mar 19 14:37:30 UTC 2007
Maarten Wiltink wrote:
> "Spoon" <devnull at localhost.com> wrote in message
> news:45fe6248$0$27199$426a74cc at news.free.fr...
>>Maarten Wiltink wrote:
>>When you say that the clock in a server next to a door will wander a
>>lot, are you referring to the effects of the sudden temperature swing
>>when a cold draft comes in the server room?
>>Does this mean that the maximum rate of change is 15 µs per second?
> No, that's something else, probably the worst-case estimate for how
> fast synchronisation deteriorates between server polls. But that's no
> more than a guess.
> The number I mean is the 500PPM cap on the observed drift that NTP will
> accept before giving up. You can't change that without recompiling, IIAMN.
> Maarten Wiltink
500PPM is the limit to the rate at which the clock can be slewed in most
operating systems. Recompiling will not help. Modifying the O/S kernel
and rebuilding the kernel MIGHT help. (I've never heard of anyone having
done this.) If the system is that badly broken it's probably easier to
have the hardware repaired.
More information about the questions