Table of Contents

About

This page is about metrics as telemetry data context (realtime performance, monitoring of system)

A telemetry metric is a time serie with:

  • timestamped
  • numerical value
  • and optional labels

Other name

Monitoring Metrics are also known as:

  • Statistical Collector
  • Perfcounter (generally on Windows)
  • Performance metrics
  • Performance data

They are known as statistical data collectors because they gather data and compute statistics. They are used to collect, evaluate and report results, e.g.

  • mean residence times of customers
  • or the average utilisation of a server.

Aggregated metrics / Statistics

An aggregated metrics is

They are event metrics.

For the business related view of metrics, see Dimensional Data Modeling - Metrics - Key Performance indicators (KPI)

Type

Counter

A Counter is a metric that:

  • can go only up.
  • is reset when the application restarts.
  • is used to record occurrences of events like the number of customer arrivals
  • is plotted with a rate function over an interval

Car example: the mileage counter.

Gauge

A Gauge is a metric that can go up and down (gauge values are not rates).

Car Example: Speedometers

Timer

A Timer measures both the count of timed events and the total time of all events timed.

Others

  • Distribution
  • Histogram - A histogram ranks the measured values into predefined classes.
  • Regressions: A Regression perform a regressional analysis of two different collected value series (x and y). It reports the number of observations, mean (for x and y), regression and correlation coefficients, standard deviations and interception.

Time series:

  • Tallies. A tally records the mean and standard deviation of a (time) series of values. Example: Computing the average time customers spent on a system
  • Accumulates: An Accumulate records the mean and standard deviation of a time series of values against different time intervals. Example: System load

See also: Prometheus Metrics Type

Implementation