sample_rate

Supported in:

Rules

optimization.sample_rate(byteOrString, rateNumerator, rateDenominator)

Description

This function determines whether to include an event based on a deterministic sampling strategy. This function returns:

• true for a fraction of input values, equivalent to (rateNumerator / rateDenominator), indicating that the event should be included in the sample.
• false indicating that the event shouldn't be included in the sample.

This function is useful for optimization scenarios where you want to process only a subset of events. Equivalent to:

hash.fingerprint2011(byteOrString) % rateDenominator < rateNumerator

Param data types

• byteOrString: Expression that evaluates to either a BYTE or STRING.
• rateNumerator: 'INT'
• rateDenominator: 'INT'

Return type

BOOL

Code sample

events:
    $e.metadata.event_type = "NETWORK_CONNECTION"
    $asset_id = $e.principal.asset.asset_id
    optimization.sample_rate($e.metadata.id, 1, 5) // Only 1 out of every 5 events

  match:
    $asset_id over 1h

  outcome:
    $event_count = count_distinct($e.metadata.id)
  // estimate the usage by multiplying by the inverse of the sample rate
    $usage_past_hour = sum(5.0 * $e.network.sent_bytes)

 condition:
  // Requiring a certain number of events after sampling avoids bias (e.g. a
  // device with just 1 connection will still show up 20% of the time and
  // if we multiply that traffic by 5, we'll get an incorrect estimate)
  $e and ($usage_past_hour > 1000000000) and $event_count >= 100