[time-nuts] Comparing the BeagleBone Black & Raspberry Pi as NTP servers

[David J Taylor]

From: Attila Kinali

Something is wrong here. I would expect the BBB to perform at least
as well as the rpi (after all, the BBB has an ethernet MAC with
IEEE1588 support, while the rpi is basically a glorified USB controller
with attached graphics card).

You are most likely running services on the BBB (network services,
local services, cron jobs,...) that cause the high, and spikey
cpu load, which in turn destroys your ntp performance.

Also, compare your results to [1], where Dan Drown uses the
capture/compare unit of the AM3359 to timestamp the PPS and use
this for ntp. In [2] he tries to measure the temperature dependence
of the BBB oscillator (not be best way, but...).

Attila Kinali


[1] http://blog.dan.drown.org/beaglebone-black-timer-capture-driver/
[2] http://blog.dan.drown.org/tcxo-beaglebone-black/
========================

Attila,

Yes, my posting was at least in part for help with resolving the much higher 
CPU load seen on a default BBB installation than on an RPi setup.  If you 
could provide a list of services and cron jobs to be disabled, and a brief 
guide to doing so (as I'm a beginner with Linux) that would be appreciated.

I had looked at [1] but I'm not using his special timer setup, just GPIO 
pins.  Unfortunately he doesn't quote any of the usual NTP parameters with 
which to compare.

I'm also measuring thing differently, in that I'm measuring from a remote 
rather than a local system, as that is what would happen when using the 
device as an NTP server.

Thanks,
David
-- 
SatSignal Software - Quality software written to your requirements
Web: http://www.satsignal.eu
Email: david-taylor at blueyonder.co.uk