[Beowulf] precise synchronization of system clocks

[Lux, James P]

On 9/30/08 8:37 AM, "Robert G. Brown" <rgb at phy.duke.edu> wrote:

> On Tue, 30 Sep 2008, Lux, James P wrote:
>
> This is a very nice response, and I think you're on a very good track.
> IIRC from discussion a few years ago, GPS can yield what, microsecond or
> better timing (if used to adjust drift and resync all clocks)?  In
> principle sub-microsecond, since a microsecond is order of 300 meters
> and GPS can get you within 30.

The grottiest grungy GPS receiver can probably do 100ns on its 1pps tick,
and most are in the 20ns range.  There ARE receivers that have systematic
errors (i.e. Some sort of sawtooth in the error) and, of course, there are
countless schemes to compensate in one way or another.

When you get into the 1ns area, you have to start worrying about multipath
(1 ft/ns and all that) and the apparent position of the antenna (phase
center for you EM folks) varying as a function of look angle (because the
satellites are moving across the sky, eh?)  Again, there are fancy
algorithms to compensate for them (because it's repeatable and predictable).
Also such things as ionospheric variations, temperature variations on the
antenna and coax cable, and hordes of other problems.

But, if you're trying to sync the CPUs in your Beowulf to 1ns, I think
you'll have bigger problems than the GPS time reference moving around.  Time
to buy yourself a good crystal oscillator.

GPS isn't all that hot in the fine scale (20ns in 1 sec is 20ppb, 2E-8) even
cheap quartz is better (<1E-10 over 1 second). What it's good for is
"absolute" calibration (so you can know that your quartz oscillator is
10.000 000 000 MHz and not 10.000 000 001 MHz) and for establishing
synchronization over a large distance (e.g. If your cluster is spread over
100km, GPS can be used to sync it to the 20ns level.  A notional application
along these lines is the cluster of computers that form the cellphone
switching system, distributed among those fake tree towers, etc.)

For more info than anyone should want to know:
http://www.ieee-uffc.org/fcmain.asp?view=review#tutor
The tutorial from John Vig is particularly to be recommended (especially for
you physicists out there.. Quartz is a fascinating material, anisotropic,
etc.)