We have started to run the network time protocol daemon ntpd some time ago, here are the daily plots of the statistics. It seems that the time offset stays in the interval [-10 10] ms, but there are some disturbances which are probably due to congestion on the link between iwasno.net and the NTP peer.
The following plot has been created locally (last night) using a gnuplot with PNG support (hence the nicer text font, I also increased the size to 800x600). Note that delay, jitter and offset values outside [-50 150] ms are not plotted to reduce the overlap between the curves.
The following plot shows the frequency offset of the old server (box7842).
There seems to some correlation between frequency error and temperature although it is not as strong as in this example.
Temperature data courtesy of the University of Duisburg-Essen ALBERT-KRATZER-KLIMASTATION (Stadtstation) refers to outside, whereas the machine is in a temperature-controlled data center.
The NTP peer changed often (usually after each reboot), this can be clearly seen as a jump in the delay and a change in the noise on the time offset, frequency offset curves.
Some information about peers used in the first few months can be found on the NTP peers page.Last changed: 2014-10-16T23:49:33-CEST.
back to Roland's start page.