Anselmo-Ed@cs.yale.edu (Ed Anselmo) (04/11/90)
Speaking of ntp -- I've brought it up here on both Sun-4/60's and
Sun-4/390's. You wouldn't think that there'd be all that much
difference between the machines, but ...
I'm using precision -7 and tickadj 5 on both machines. The 4/60's
keep great time, but the 4/390's seem to have consistently greater
offsets -- the difference is about an order of magnitude. Plus the
4/390's clutter up the console with silly warning messages whenever
the time is reset.
From a 4/60:
(rem) Address (lcl) Strat Poll Reach Delay Offset Disp
==========================================================================
*CF-GW.CS.YALE.E 128.36.16.2 2 64 377 20.0 -7.0 2.0
From a 4/390:
(rem) Address (lcl) Strat Poll Reach Delay Offset Disp
==========================================================================
-CF-GW.CS.YALE.E 128.36.0.21 2 64 376 20.0 188.0 127500.0
-ZOO-GW.CS.YALE. 128.36.0.21 3 64 376 30.0 91.0 63500.0
Would a different value for tickadj and/or precision work better with
the 4/390's?
--
Ed Anselmo anselmo-ed@cs.yale.edu {harvard,decvax}!yale!anselmo-edMills@udel.edu (04/11/90)
Ed, I can't help you a lot with the platform mechanics, but I do observe the dispersion on the 4/390 is outta sight. The precision parameter should not affect this, but the tickadj might; however, I would suspect the culprit is lost clock interrupts. Others on this list might be able to give more definitive anecdotal evidence. Ahem. Dave