noise

README

Noise Processor Plugin

The Noise processor is used to add noise to numerical field values. For each field a noise is generated using a defined probability densitiy function and added to the value. The function type can be configured as Laplace, Gaussian or Uniform. Depending on the function, various parameters need to be configured:

Configuration

# Adds noise to numerical fields
[[processors.noise]]
  ## Specified the type of the random distribution.
  ## Can be "laplacian", "gaussian" or "uniform".
  # type = "laplacian

  ## Center of the distribution.
  ## Only used for Laplacian and Gaussian distributions.
  # mu = 0.0

  ## Scale parameter for the Laplacian or Gaussian distribution
  # scale = 1.0

  ## Upper and lower bound of the Uniform distribution
  # min = -1.0
  # max = 1.0

  ## Apply the noise only to numeric fields matching the filter criteria below.
  ## Excludes takes precedence over includes.
  # include_fields = []
  # exclude_fields = []

Depending on the choice of the distribution function, the respective parameters must be set. Default settings are noise_type = "laplacian" with mu = 0.0 and scale = 1.0:

Using the include_fields and exclude_fields options a filter can be configured to apply noise only to numeric fields matching it. The following distribution functions are available.

Laplacian

• noise_type = laplacian
• scale: also referred to as diversity parameter, regulates the width & height of the function, a bigger scale value means a higher probability of larger noise, default set to 1.0
• mu: location of the curve, default set to 0.0

Gaussian

• noise_type = gaussian
• mu: mean value, default set to 0.0
• scale: standard deviation, default set to 1.0

Uniform

• noise_type = uniform
• min: minimal interval value, default set to -1.0
• max: maximal interval value, default set to 1.0

Example

Add noise to each value the inputs.cpu plugin generates, except for the usage_steal, usage_user, uptime_format, usage_idle field and all fields of the metrics swap, disk and net:

[[inputs.cpu]]
  percpu = true
  totalcpu = true
  collect_cpu_time = false
  report_active = false

[[processors.noise]]
  scale = 1.0
  mu = 0.0
  noise_type = "laplacian"
  include_fields = []
  exclude_fields = ["usage_steal", "usage_user", "uptime_format", "usage_idle" ]
  namedrop = ["swap", "disk", "net"]

Result of noise added to the cpu metric:

- cpu map[cpu:cpu11 host:98d5b8dbad1c] map[usage_guest:0 usage_guest_nice:0 usage_idle:94.3999999994412 usage_iowait:0 usage_irq:0.1999999999998181 usage_nice:0 usage_softirq:0.20000000000209184 usage_steal:0 usage_system:1.2000000000080036 usage_user:4.000000000014552]
+ cpu map[cpu:cpu11 host:98d5b8dbad1c] map[usage_guest:1.0078071583066057 usage_guest_nice:0.523063861602435 usage_idle:95.53920223476884 usage_iowait:0.5162661526251292 usage_irq:0.7138529816101375 usage_nice:0.6119678488887954 usage_softirq:0.5573585443688622 usage_steal:0.2006120911289802 usage_system:1.2954475820198437 usage_user:6.885664792615023]

Documentation

Index

• type Noise
  • func (p *Noise) Apply(metrics ...telegraf.Metric) []telegraf.Metric
  • func (p *Noise) Init() error
  • func (*Noise) SampleConfig() string

Types

type Noise

type Noise struct {
	Scale         float64         `toml:"scale"`
	Min           float64         `toml:"min"`
	Max           float64         `toml:"max"`
	Mu            float64         `toml:"mu"`
	IncludeFields []string        `toml:"include_fields"`
	ExcludeFields []string        `toml:"exclude_fields"`
	NoiseType     string          `toml:"type"`
	Log           telegraf.Logger `toml:"-"`
	// contains filtered or unexported fields
}

func (*Noise) Apply

func (p *Noise) Apply(metrics ...telegraf.Metric) []telegraf.Metric

func (*Noise) Init

func (p *Noise) Init() error

Creates a filter for Include and Exclude fields and sets the desired noise distribution

func (*Noise) SampleConfig

func (*Noise) SampleConfig() string