Performance seen from 2 CERN PingER Monitoring Stations

(click on any image to zoom in)	For many sites the RTTs are close and changes track well. A slight difference can be understood as the pings are probing the network at slightly different times, so the close agreement suggests performance between CERN and these sites is quite constant.
(click on any image to zoom in)	For other sites the RTT do not appear to agree very well, but this can be understood if the network between CERN and these sites is varying over a short timescale, and the time difference between the pings means different network conditions are being probed.
(click on any image to zoom in)	Many sites have large differences in RTT, but these differences appear to be constant. In all cases the longer time is from suncs02, the machine inside the fierwall, so one scenario is the firewall is adding extra milliseconds to the RTT for these sites. If this is the case why does the firewall not effect the other sites ?
(click on any image to zoom in)	Some sites have large differences in RTT, but they also do not appear to track very well. This I find very odd.
(click on any image to zoom in)	The effect continues into January. Although the data is somewhat patchy for January, the same pattern clearly exists.