Graphite is an enterprise-scale monitoring tool that runs well on cheap hardware. It was originally designed and written by Chris Davis at Orbitz in 2006 as side project that ultimately grew to be a foundational monitoring tool. In 2008, Orbitz allowed Graphite to be released under the open source Apache 2.0 license. Since then Chris has continued to work on Graphite and has deployed it at other companies including Sears, where it serves as a pillar of the e-commerce monitoring system. Today many large companies use it.
What Graphite does not do is collect data for you, however there are some tools out there that know how to send data to graphite. Even though it often requires a little code, sending data to Graphite is very simple.
Feeding in your data is pretty easy, typically most of the effort is in collecting the data to begin with. As you send datapoints to Carbon, they become immediately available for graphing in the webapp. The webapp offers several ways to create and display graphs including a simple URL API that makes it easy to embed graphs in other webpages.
Graphite consists of 3 software components:
A Twisted daemon that listens for time-series data.
Data collection agents connect to carbon and send their data, and carbon's job is to make that data available for real-time graphing immediately and try to get it stored on disk as fast as possible.
Carbon is made of up three processes:
The primary process is carbon-agent.py, which starts up the other two processes in a pipeline. Carbon-agent accepts connections and receives time series data in the appropriate format. This data is sent through the pipeline to carbon-cache, who stores the data a cache where data points are grouped by their associated metric. Carbon-cache constantly attempts to write the largest such group of data points down the pipeline to carbon-persister. Carbon-persister reads these data points and writes them to disk using whisper.
The reason carbon is split into three processes is actually because of Python's threading problems. Originally carbon was a single application where these distinct functions were performed by threads, but alas Python's GIL prevents multiple threads from actually running concurrently. Since the initial deployment of Graphite was done on a machine with lots of rather slow CPU's, we needed true concurrency for performance reasons. Thus it was split into three processes connected via pipes.
Whisper is a fixed-size database, similar in design to RRD (round-robin-database). It provides fast, reliable storage of numeric data over time.
RRD is great, and initially Graphite did use RRD for storage. Over time though, we ran into several issues inherent to RRD's design.
A Django webapp that renders graphs on-demand using Cairo.
Upon receiving a rendering request, the Graphite webapp simultaneously retrieves data for the requested metrics from the disk and from carbon's cache via a simple cache query socket that carbon-cache provides. Graphite then combines these two sources of data points into a single series, which is then rendered. This ensures that graphs are always real-time, even when the data hasn't been written to disk yet.