Monitoring Taxonomy

Chapter 3. traditional.free_open.collectors.data

This is a lonely category. Most monitoring collectors focus on system metrics like CPU, disk, and memory, but the tools in this path are general-purpose data collectors.

StatsD: Simple Daemon for Stats Aggregation

StatsD is a cache-and-forward metrics summarization daemon. It listens for metrics on a network socket, holds/aggregates them for some amount of time (10 seconds by default), and then emits a summary of the data (percentiles, sums, averages, etc.) to one or more (pluggable) backend metrics services like Graphite or Librato.

StatsD is also a vast collection of clients for the StatsD server, which make it trivial to emit metrics into a listening StatsD interface from pretty much anything. These include command-line tools as well as integrations for other monitoring systems and language bindings for myriad programming languages. StatsD is pretty much the de facto means of wiring together a metrics-centric toolchain today.

Push, Pull, Both, or Neither?

StatsD is a push-based system.

Measurement Resolution

The standard tick is 10 seconds.

Data Storage

StatsD is in-memory cache-and-forward: there is no data-storage requirement.

Analysis Capabilities

None. StatsD is intended to be pointed at the analysis tool of your choice.

Notification Capabilities

None. Notification typically occurs before StatsD (at a monitoring system like Nagios) or after StatsD hands off the metrics (in a processing system like Riemann or Graphite/Bosun) ...