NTP?

Greg Lindahl lindahl at conservativecomputer.com
Thu Oct 11 15:37:19 PDT 2001


> > Now if you want GOOD time synchronization (to within a microsecond)
> > that's a *much* harder problem...
> 
> Yes, it starts to become very difficult and expensive right in there
> somewhere.  The so called "atomic clocks" one can now buy in stores like
> Wal Mart for $50 or so are autosync'd to WWVB at NIST in Colorado.

Um, you're off in the wrong direction. I was thinking of machines like
the Cray T3E, which weren't synched to the rest of the world any
better than anyone, but if you called gettimeofday() on different cpus,
you'd get a *good* answer, but a relative one.

Second, I've used the radio clocks that you mention, and they aren't
called "atomic clocks", and you CAN compensate for the 3 microseconds
because you know the propagation speed of radio waves in the
atmosphere.  But it actually has fairly nasty dispersion. That's why
everyone (including the currently NTP layer 0 hosts) uses GPS clocks,
which you also mention.

When I used to work in midtown Manhattan, we had a building blocking
our western view, so we got the signal bounced off a convenient other
building.

greg




More information about the Beowulf mailing list