A java library to collect metrics about a software, especially servers, during run time and send it to a central system such as StatsD or graphite.
The Metrics library written by Codahale while working at Yammer has gained a strong following in the Java server development community. This library was novel at its inception in that it aggregated metrics on the server node itself and then sent pre-aggregated metrics a central system such as graphite or statsd. This allows you to use the spare CPU cycles and memory for aggregation. Most application servers are not run to capacity and hence there is a lot of available free resources. It would be more expensive to run the same aggregation on a central node or cluster. Pre-aggregation also saves on network bandwidth as the raw metrics could be generated at a very high rate. With on-node aggregation all the metrics are aggregated in memory locally and only shipped to central systems periodically.
In addition the codahale library allows you to easily integrate application specific metrics in addition to more general server metrics that would be captured by traditional systems such as CollectD or JMX.
There are ports of codahale's metrics Libraries in several languages including:
