[cgl_discussion] [Timers evaluation/Validation] Notes from Ti mers eval team meetin g

Craig Thomas craiger at osdl.org
Wed Jan 22 14:33:51 PST 2003


I was thinking on a much smaller granularity than NTP.  If we need to
measure milliseconds or microseconds, NTP may not be the best way.
I am unfamiliar with what can be done, so I don't know the answer here.

I was just thinking out loud...


On Wed, 2003-01-22 at 13:50, Fleischer, Julie N wrote:
> > Craig Thomas wrote:
> > I don't mean to make a joke of this (I am really serious), but how do
> > you time a timer?  This will be a hard problem to solve.  Trying to
> > determine if timers respond in an acceptable fashion under 
> > heavy stress
> > would mean that you need to make sure that they keep their time.  But
> > how to you measure how they keep their time?  
> > 
> > With another set of (non posix) timers, perhaps?
> 
> It's good that you bring this up.  My logic was (and feel free to point out
> holes as some of it was just assumptions I didn't think about questioning
> yet):
> 
> - For timers, they need to function relative to a clock.  Even if the clock
> is off, if the timer expires at the correct time relative to that clock, I
> would consider that correct behavior.  So, I can "time a timer" with the
> clock used for the timer.
> 
> - For clocks, we'll have to use an external known-good clock somehow (Like
> maybe run clock tests and compare time via NTP.).
> 
> - Julie
> 
> **These views are not necessarily those of my employer.**
>  
> _______________________________________________
> cgl_discussion mailing list
> cgl_discussion at lists.osdl.org
> http://lists.osdl.org/mailman/listinfo/cgl_discussion
-- 
Craig Thomas <craiger at osdl.org>
OSDL




More information about the cgl_discussion mailing list