Brad Knowles wrote:
At 10:46 AM -0500 2005-09-09, wayne wrote:
Links that cross oceans tend to have high jitter and high latency.
Ending up with an NTP server that is half way around the world is
going to cause you to have something like a 100ms minimum delay.
(65ms for the speed of light, but electric signals only travel at
2/3c, plus router delays, etc.)
Then links that go across the ocean once and then come back across
the ocean, in order to get from one country to the one next door,
would be at least twice as bad. Yet this is still the case today
amongst many sites in Asia -- they have really, really bad
connectivity with each other, and much better connectivity with the
US. Until relatively recently, if you wanted to go from Korea to
Japan, all those packets would be routed via San Francisco.
Can we stop nerding? And ask ourselves what our users expect from a pool
server? If an ordinary user starts using ntpd, (or ntpdate) why does he
do it: He's sick of setting his computer clock every week. Most computer
users are happy if their system clock is less than five minutes off. I
like my time a bit better, but would settle for a system that promises
synchronisation with UTC with only a few seconds of error.
Most of the pool users use NTP to prevent their computer clock from
skewing too much, the minute must be right, and if the seconds are
correct, well that's nice. You are talking about the effects of 100ms
latency (~10 ms jitter...) 98% of our users could care less. The ones
that do care can/will/should take the time to find a set of ntp servers
that suits them.
Greetings,
Peter.
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers
