From OpenNMS

Using SNMP, JMX, HTTP, and the Nagios "NSClient" agent (to collect Windows PerfMon counters), OpenNMS can collect and store all manner of interesting data. The out of the box configuration includes numerous commonly available and useful SNMP OIDs (for the SNMP collector), and PerfMon counters (for NSClient), which will be automatically collected if found to be available on monitored equipment. Additions or other customisation of collected values is by way of easily edited XML files.

Data is stored in RRD or JRobin format; both are "Round-robin database" formats which store a fixed number of datapoints in a round-robin fashion providing consistent data storage size (unchanging for a given data point over time), and aggregation for older data to minimise data storage requirements

Of course, all this data is useless unless it can be viewed in an appropriate fashion. Graphing uses standard RRDTool style graphs (provided by JRobin graph facilities if using JRobin for storage). Out of the box configuration includes graphs for the standard collected data points. Graphs can be customised, and new graphs by editing of a simple format text file. Ad-Hoc graphs can be created using the web interface, allowing specific one-off views to be created as required.

Response times of various services are also collected and stored in the same manner, and can be viewed in graph form. This can be useful in determining responsiveness of services, possibly warning of building performance issues.

Thresholds can be set on these collected values, which if exceeded, generates events and possible notifications. Using this, situations such as extended raised CPU usage, or limited free disk space, can be detected and notified