performance, was Re: [LinuxPPS] patch report

[gnu not unix]

(waves) Hi from hot California Udo--

In message <44B9D50F.4030705 at xs4all.nl> you write:
>Hello Steven,

>gnu not unix wrote:
>> I've implemented the linuxpps setup on one of the hosts
(...)
>> http://www.wraith.sf.ca.us/ntp/rrd/index.html

These plots are a bit messed up at the moment--there was a
five hour long power failure in the area. I have backups...

>Thanks for posting!
>How do you think the code performs?
>Tips, hints, suggestions?

The combination of the new linux kernel (2.6) and the new
pps setup seem to be working fine in the limited time I've watched
it so far. The RRD plots as I've programmed them are very sensitive
to any spasms that occur. They are setup to show how bad the
ntpd is operating, in that the dynamic resizing will capture
the peaks and scale appropriately. Of course when things are
great, the offset plot often drops down to a scale of +/- 2 
microseconds, which is pretty fantastic performance.

One thing that I have noticed, is that the ppm drift of the
motherboard RTXO on the unit I switched from 2.4.20-NANO
(PPSKit) to 2.6/linuxpps, has gone from 70ppm to 50ppm.

This is very interesting and I am trying to understand the
change. The old kernel ran at 100hz tick rate and the new
kernel is setup for 250hz tick. I recall Dave Mills has
said the kernel code design shouldn't be changed, as it
is part of the software PLL, and I think that he suspects 
the linux kernel mongers of Not Paying Attention. 
On the other hand it seems to be working here, so far.
I haven't run 2.6 in production until now, however.

I might try another linuxpps with hz=100 and see what I get.
For the moment, I'm letting the setup run so I can get a
clean set of plots again. I've linked in my scripts to
the above mentioned page so if anyone else wants, they
can grab them (they're Public Domain).

../Steven