[ntp:questions] Loopstats jitter field mostly zero?

gabs gsricalde at gmail.com
Tue Nov 27 16:24:14 UTC 2012


On Sunday, November 25, 2012 3:56:28 AM UTC+8, David Taylor wrote:
> I've just started the statistics collection on my Linux/RPi NTP server, 
> and when plotting the jitter results I found that most of the values 
> were zero.  I don't see that on my Windows stratum-1 systems.  Is there 
> a limit in the calculation of the jitter value recorded (to 9 decimal 
> places) in the loopstats file.
> 
> I really can't imagine that such a low-powered server is /that/ good - 
> better than a nanosecond.
> -- 
> Cheers,
> David
> Web: http://www.satsignal.eu
>

The displayed jitter in loopstats or the peer list can be different from the
actual jitter of the source. This case is the reverse: low actual jitter,
15 us displayed jitter.

root at OpenWrt:/tmp# tail /tmp/ntp/loops
56257 76888.554 -0.000000080 6.553 0.000015259 0.000053 3
56257 76896.554 -0.000000098 6.553 0.000015259 0.000052 3
56257 76904.554 -0.000000071 6.553 0.000015259 0.000051 3
56257 76912.554 -0.000000061 6.553 0.000015259 0.000049 3
56257 76920.554 -0.000000091 6.553 0.000015259 0.000049 3
56257 76928.554 -0.000000008 6.553 0.000015259 0.000046 3
56257 76936.554 -0.000000048 6.553 0.000015259 0.000043 3
56257 76944.554 -0.000000014 6.553 0.000015259 0.000041 3
56257 76952.554 0.000000039 6.553 0.000015259 0.000038 3
56257 76960.554 0.000000016 6.553 0.000015259 0.000036 3

This is normal for this platform since ntpd seems to use the precision as the
lower bound for reported jitter. The precision is low as it takes 12 us to read
the clock and convert to NTP timestamp format.

For the RPi, I have no idea but I'll be getting one before the year ends :)



More information about the questions mailing list