[LinuxPPS] Recommendation / Re: Stats & architectures

clemens at dwf.com clemens at dwf.com
Thu Mar 13 18:54:49 CET 2008


> Rodolfo Giometti wrote:
> > I suppose you shouldn't need to much efforts in getting the HARDPPS
> > support from Ulrich Windl's code and move it into LinuxPPS.
> 
> What benefit(s) does HARDPPS support bring?
> 
Well, you would expect that having the kernel itself detect the time 
differences
from its clock to the pps signal would result in a 'tighter' control of the 
clock
than the ntpd (userspace) code doing this and then passing this information to
the kernel using the adjtime system call.  

With a clock with (only) microsecond resolution, and an adjtime system call 
with only microsecond resolution one has to expect that the clock frequency/
phase is zig-zagging around the 'correct' value a lot more than if these were
both configured in nanoseconds.  The HARDPPS support would put the entire
ntp calculation in the kernel, thus at least the adjtime system call is no 
longer
relevant.

However, there is the following comment at the end of the "The Nanokernel"
paper by Mills and Kamp:

    While the current minimum daemon update interval is currently limited to
    16s in the interest of minimizing kernel overhead, it might be acceptable 
in
    fast machines to reduce that interval to 1s.  Should this be done, it 
would be
    practical to do almost all discipline loop processing in user space and 
move
    the per-second processing to the daemon, where more flexible processor and
    memory resource commitments are possible.

I read that to say that with the fast machines we have today, that HARDPPS
is less useful than it used to be.  That a minor modification to ntpd would 
allow
the same time resolution using the usertime ntpd calculation, and minimal
kernel modification.  This would PROBABLY require a nanosecond clock
(and timecounters or whatever) and a nanosecond adjtime, but I think that
would be a good modification what with the current machine speeds.

Hope that helps.
-- 
                                        Reg.Clemens
                                        reg at dwf.com





More information about the LinuxPPS mailing list