[ntp:questions] Re: ntp client over satellite and no CMOS battery

David J Taylor david-taylor at blueyonder.co.not-this-bit.nor-this-part.uk.invalid
Sun Sep 11 10:30:20 UTC 2005


David Woolley wrote:
> In article <wAQUe.106462$G8.19768 at text.news.blueyonder.co.uk>,
> David J Taylor <> wrrote:
[]
>> In general, all these last three systems keep within 10-20ms of
>> correct time, and have only been worse than that across a power
>> glitch or across a
>
> My experience is that loaded (S at H) NT systems jitter about 20ms, but
> the same machine running Linux will have 90 percentile offsets around
> 3ms.  (My main speculation is that NT will not pre-empt a running
> process until the next clock interrupt.)

I have recently bought the book: Windows Internals (4th edition), which 
includes all of this level of detail.  I'm fairly certain that a running 
process can be pre-empted at any time when a higher priority software or 
hardware interrupt arrives, and that Windows isn't limited to just the 
times ticks to reschedule processes.

Cheers,
David 





More information about the questions mailing list