NTP primary clock data.

General information

MO Technologies network operates two stratum 1 clocks disciplined via GPS and PPS methods. Additionally, two secondary clocks are operated at a stratum 2 disciplined via PPS. This site is the equivalent of the Dave Mills backroom for Alcatel-Lucent ESD, as such we peer with machines on that network. The clock operational statistics are sampled 5 minute intervals and subsequently entered into an RRD style database for plot generation. Plots for the last 25 hour period and the last 8 days are displayed. The following information is gathered from each clock: precision, delay, dispersion, offset, jitter, noise, stability, frequency, stratum, and peer (essentially the ouput from an ntpq -c rv command).

Additional clocks are maintained for Alcatel-Lucent ESD. These are tracked here for referece.

At some point I would like to gather the raw data from each GPS unit and track the Allan deviation and other interesting data. This will come about from the generous folks on the Time-nuts list; however, until I get motivated to write come code and build an interface, this data shall remain trapped in the Z8316A and Z8301 units.....

Additional NTP clocks are operated for specific client groups. These clocks are slaved to the primary clock. (and at least 9 external clocks as well). Operational statistics are gathered for these clocks as well. bureau; credenza; The Firewall. Client stats can be found here

This data is gathered using scripts shamelessly stolen from Steven Bjork. (and then highly modified by me.) The ksh script, do-ntp-rrdstat, collects the data via mode 7 ntpq queries and updates an RRD data file with them. The actual images are generated by the do-rrdntp script.

Raw NTP statistics are gathered from the loopstats, peerstats, and sysstats files. This data is then plotted using GNUplot. Because this data is gathered from the NTP timing loop, it will contain a more accurate picture of NTP performance than the high level data that is sampled every 5 minutes (by do-rrdntp). Servers for which we have both 5 minute RRD sampled data and NTP timing loop data, are indicated with the green hyperlink.

For comments about this page contact Mark Owens.