Skip to content

Glossary

Quick Lookup

Common terms used in this module:

  • Signal - a derived metric created from raw data
  • Raw Metric - a directly recorded measurement from a system
  • Derived Metric - a value calculated from one or more raw metrics
  • Error Rate - proportion of errors relative to total requests
  • Latency - the time required for a system to respond to a request
  • Throughput - the volume of work processed by a system (for example, number of requests)

Raw Metric

A measurement recorded directly from a system. Examples:

  • number of requests
  • number of errors
  • total latency

Raw metrics describe what happened but may not be the easiest values to interpret.

Signal

A signal is a derived metric designed to make system behavior easier to understand. Signals are usually created by combining or transforming raw metrics. Examples:

  • error rate
  • average latency per request
  • throughput

Signals help analysts monitor system health and detect unusual behavior.

Error Rate

The proportion of failed requests relative to total requests. This signal helps identify reliability problems in a system.

error_rate = errors / requests

Latency

The time required for a system to respond to a request. In this module, we often compute average latency per request:

avg_latency = total_latency_ms / requests

Throughput

The amount of work processed by a system in a given observation period. Examples include:

  • requests handled
  • transactions processed
  • messages delivered

Feature Engineering

The process of creating new variables (features or signals) from existing data. Signal design is a form of feature engineering for monitoring systems. Feature engineering often includes:

  • ratios (errors / requests)
  • averages (latency per request)
  • aggregations (rolling means)
  • transformations (log, scaling, normalization)

Signal Design

The process of creating useful signals from raw system metrics. Well-designed signals make system behavior easier to monitor and interpret, which is essential for continuous intelligence systems.