tl;dr click here

When we talk about capturing metrics in applications. One server/service that constantly is in all conversations monitoring, is statsd. Incase you have never heard of it, statsd is a udp/tcp server that you send your in-code metrics to. These metrics get aggregated by statsd, and are forwarded to various backends. Some backends are services like librato or sumologic. Other times you are sending metrics to time series databases such as graphite or god forbid influxdb.

This boils down to in code you can say "log whenever this block of code is hit" or say "measure how long this function takes to execute". These stories come together to form pretty graphs, and rich alerts. All of this enabled by statsd.