[cadynce] NTPMeasurements < CADYNCE < TWiki

[Gautam Thaker]

Folks:

I have done some NTP measurements at Emulab and have put up a wiki page on it. Please see link below.

It appears that if NTP is allowed to run for a few hours after reboot time difference between two machines should be 1-2 msec at 
most. I believe it can get sub msec if allowed to run even longer.

Our earlier reported problems with clock sync appear due to:

1) not letting Emulab nodes settle for long enough
2) we now believe that the RT kernels we use may need a newer/recompiled ntpd. We are investigating this further.

In anyevent, at VU Isislab or at UMD lab where machines don't get reboot all the time things should be quite good.

https://escher.isis.vanderbilt.edu/twiki/bin/view/CADYNCE/NTPMeasurements

Gautam