[ntp:questions] follow up on Windows 2003 SP1 Time Service versus ntpd

Ry malayter at gmail.com
Wed May 24 22:20:44 UTC 2006

I've finally gotten around to posting a chart showing the results of my
tests of the improved Windows 2003 SP1 time service:

As you can see, it seems to keep time roughly within 16 ms or so of
ntpd. This seems in-line with the precision the Windows Time Service
reports: 2^(-6). I believe the saw tooth pattern is quantization noise
resulting from this low precision.

I haven't had time to feed my stats files into anything more
advanced... are there any pre-baked ntp logfile analysis tools out
there that you might suggest?

For the record, here's a description of the test methodology:

Two similar servers on the same gigabit LAN were loaded with Windows
2003 SP1. One system had the Meinberg distribution of ntpd 4.02b
installed; the other was left with the default Windows Time Service.
Both were configured to use the same five stratum-1 time servers, and
started at the same time. The ntpd server monitored the other server
for a week using "server x.x.x.x noselect" in its config file, with
results logged in peerstats and loopstats.

More information about the questions mailing list