While I think
cricket is a great idea (in theory), in practice I found doing the simplest monitoring difficult. In fact, I couldn't get it to monitor eth0 on a linux box. I burned 2-3 hours trying the various contributed packages for monitoring eth0 with net-snmp, I even tried to work through my own configuration to do this, but I found the documentation lacking in working examples. Finally I gave up and went back to good old
MRTG . MRTG is steady as a rock. When you don't have a huge number of devices to monitor, nothing works better. I set up the monitoring I wanted in 20 minutes. Cricket was basically created for people needing to monitor hundreds of devices, and to handle the case were the devices don't have the same SNMP community string, which MRTG doesn't handle well. I think if I was trying to monitor a switch or router that cricket would have worked beautifully, but it had trouble monitoring a server.
I went ahead and used the RRD database format so that I can export my bandwidth data to XML. Switching MRTG to the RRD format means you need
mrtg-rrdtool.cgi to draw the graphs for you.
