When building distributed systems, tools like Inca, Monit and M/Monit help you monitor status and events, for example software failures, network dropouts and much more. However, they are not designed to monitor the network links themselves, even though they are of equal importance.

Today we've installed and configured an instance of the open-source Smokeping program on our servers to monitor several key links and systems. It works by sending 20 56-byte pings every five minutes, and plotting the results on multiple time scales. At 1120 bytes every five minutes, the added load is miniscule, and the graphs clearly show congestion, jitter and any problems. We find it to be a vary useful tool.

View it at http://iguassu.sdsc.edu/cgi-bin/smokeping.cgi and let us know what you think.