Skip to content

Glossary

Jacek Masiulaniec edited this page Apr 8, 2015 · 115 revisions

Key concepts are in bold.

aggregator

Central service acting as a publisher of a feed that carries data points pertaining to all hosts (union of forwarder and poller feeds).

controller

Central service responsible for cluster management.

data point

Number representing reading of a statistic at a specific time. Element of a time series. Produced by plugins.

In a line plot, it corresponds to every (x,y) that contributes to line's shape.

feed

TCP connection carrying a real-time stream of recently extracted data points. Produced by forwarder, poller, and aggregator.

filter

Ruleset used to drop data points received from plugins, or to rewrite them before publishing to subscribers.

forwarder

Agent distributed to every host. Publisher of a feed that carries data points pertaining to the local host.

plugin

User program running a loop that extracts data points from the local host (if executed by forwarder) or from remote hosts (if executed by poller). It submits these points to forwarder/poller for filtering and delivery to subscribers.

poller

Central service acting as a publisher of the feed that carries data points pertaining to remote hosts (e.g. SNMP devices).

statse

Plugin responsible for deriving data points from event streams (e.g. web logs) through continuous statistical aggregation.

subscriber

User program that consumes the feed.

  1. An indirect subscriber receives data points from aggregator.
  2. A direct subscriber receives points from forwarders and the poller.

time series

Sequence of data points created by periodically reading the value of a statistic (e.g. cpu utilization of the local host).

In a line plot, it corresponds to the line itself.

time series identifier

Human-readable name of a time series.

The identifier consists of two sections:

  1. Metric: A dot-delimited string that represents a broad category of time series (e.g. os.filesystem.usage).
  2. Tags: List of key=value pairs that subdivide metric along independent dimensions (e.g. host=localhost,mount=/var/log).

In a line plot, it corresponds to line's label in the legend (key) section.

Clone this wiki locally