On Wed, Mar 04, 2009 at 05:59:30PM +0000, Sam Mason wrote: > I seem to have chucked all my data away, but I'll get my logging going > and see if I can put something like this together again. I think I > found the best was to dump out the binary stream of the incoming NTP > "response" packets. I can't remember how I identified them as being > response packets now, but it shouldn't be too hard.
I seem to have recorded outgoing NTP packets; I've got about three hours of data now and have done some cheesy MS Excel style plots of the data: http://samason.me.uk/~sam/tmp/ntp-pps-timeseries.png The plot on the top shows counts of the raw packets as they're being sent out, middle one shows the distribution within a second, and the one on the bottom the distribution over a minute. I'd interpret them as follows; the spikes on the top plot show when I enter and leave the pool (@, europe and uk pools) but I'm not sure why spikes are such different shapes. I'd expect the middle plot to exhibit less heterogeneity than it appears to, i.e. I'm not sure why I'm sending more packets at 0.3, 0.4 and 0.8 seconds. When I looked in more detail this pattern appeared to stay reasonably stable over time, but could just be an artifact from the small quantity of data I've received. The bottom plot is the interesting one, it shows that there are strong biases in getting the time on the minute. I'd assume that the fact the spike shows up at 2 seconds past the minute is to do with DNS resolution. Not sure what to attribute the other fluctuations to! -- Sam http://samason.me.uk/ _______________________________________________ timekeepers mailing list [email protected] https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers
