[LinuxPPS] Re: strange readings...

[Udo van den Heuvel]

Rodolfo Giometti wrote:
> On Thu, May 17, 2007 at 01:40:30PM -0700, Martin wrote:
>>
>> Maybe it is the switch to 2.6.21...
>>
>> I know almost nothing about kernel internals, but I just
>> read somewhere that:
>>
>> "As of Linux kernel version 2.6.21, the kernel no longer has
>> a fixed 1000Hz timer tick."
>>
>> Maybe this has implications for ntp? Anyone have any thoughts?
> 
> Regarding to LinuxPPS it should have no implications at all since it
> doesn't use HZ nor it requires a fixed value.

Then how do we explain this *very* *bad* perfomance?
Booting back into 2.6.20.7 (and installing ntp 4.2.4) immediately gives
me the previous and good performance.
See the graphs at http://pindarots.xs4all.nl/mrtg/ntpstat.html to see
the effects and the small period in week 20 when I booted back into
2.6.20.7.