New Relic APM Practitioner key terms

Any monitored thing New Relic tracks as a distinct object, such as an application, host, browser app, or service. Entities have a GUID, a type, and tags, and are the unit that alerts, dashboards, and relationships reference.

A globally unique identifier assigned to each entity. It stays stable across the entity's lifetime, so it is used in APIs, dashboards, and alert definitions to point at one specific monitored object without relying on its display name.

Language-specific library installed inside your application that instruments requests, database calls, and external services, then sends the collected telemetry to New Relic. Agents exist for Java, .NET, Node.js, Python, Ruby, PHP, and Go, among others.

A single unit of work the agent times end to end, typically one web request, message consumption, or background job. Transactions are grouped by name and roll up into throughput, response time, and error-rate metrics.

A detailed breakdown of one slow transaction, showing time spent in each segment such as method calls, database queries, and external HTTP calls. Traces are sampled, so only a subset of slow requests is captured for inspection.

A timed span inside a transaction trace representing one operation, like a database query or an outbound HTTP call. Nested segments show where time is spent, making it possible to pinpoint the exact slow call in a request.

A satisfaction score from 0 to 1 based on a target response time T. Requests faster than T are satisfied, those between T and 4T are tolerating, and slower ones are frustrated. The formula is (satisfied + tolerating/2) / total.

The rate at which an application or transaction handles work, usually expressed as requests per minute (rpm) or calls per minute (cpm). It is a core APM chart metric and a common basis for alert conditions.

A technique that follows one request across multiple services by propagating a trace ID and parent span ID through headers. It stitches together spans from each service so you can see the full path and find the slow hop.

A single timed operation within a distributed trace, with a name, start time, duration, and parent relationship. Spans are the building blocks of traces and are stored in the Span event type, queryable with NRQL.

A unique identifier generated at the start of a distributed trace and passed to every downstream service. Filtering by trace ID in logs, spans, or errors lets you reconstruct one request's complete journey across systems.

A visual graph of entities and their call relationships, built from observed traffic. Nodes are services, databases, or queues, and edges show dependency direction and health, helping you spot a failing upstream or downstream dependency.

Real user monitoring that uses a JavaScript agent in the page to measure load timing, AJAX calls, JavaScript errors, and user interactions in actual browsers. It reveals front-end slowness that server-side APM cannot see.

Google's user-experience metrics: Largest Contentful Paint for loading, Interaction to Next Paint for responsiveness, and Cumulative Layout Shift for visual stability. New Relic browser monitoring captures these from real user sessions.

A period of user activity in a browser or mobile app, typically ending after a period of inactivity. Sessions group page views, interactions, and errors, and are the basis for session traces and replay features.

A sampled record of everything that happened during one browser session, including page loads, AJAX calls, and JavaScript errors. It helps reproduce a user-reported problem by replaying the sequence of events.

Scripted checks that run from managed locations on a schedule to test availability and performance. Simple checks ping a URL; scripted browsers run multi-step user journeys, catching outages even when no real users are online.

Agent-based monitoring for iOS and Android apps that captures crashes, network requests, and screen-level performance. It shows how app behavior and device or network conditions affect real mobile users.

New Relic's set of AI-assisted capabilities that detect anomalies, correlate incidents, and suggest likely root causes. It surfaces unusual behavior across services and infrastructure so responders can investigate faster than manual chart review.

Automatic identification of behavior that deviates from a learned baseline, such as a latency spike at an unusual hour. It reduces reliance on fixed static thresholds that fail to match normal traffic patterns.

An alert condition that compares current values against a learned seasonal baseline rather than a fixed number. It fires when behavior departs significantly from what is normal for that time of day or week.

New Relic Query Language, a SQL-like syntax for querying telemetry. Statements start with FROM a data type and support SELECT, WHERE, FACET, TIMESERIES, and aggregation functions, and power dashboards, alerts, and APIs.

A named table of telemetry in the New Relic database, such as Transaction, Span, Log, or Metric. Each event type has its own attributes, and NRQL queries must specify one as the source after FROM.

A key-value pair attached to an event, span, log, or metric, such as http.statusCode or appName. Attributes can be queried, faceted, and used to filter, and custom attributes can be added to enrich telemetry.

A shareable collection of charts and widgets, usually NRQL-driven, that displays key metrics for a team or service. Dashboards can be filtered by tags and used for operational reviews or incident triage.

A container that groups alert conditions and defines incident preferences, such as severity and whether to open one issue per condition or per entity. Policies are applied to entities through their tags or explicit selection.

A rule that evaluates a metric or NRQL query against a threshold and triggers when breached. Conditions define the signal, threshold, evaluation window, and whether the breach must persist before firing.

The record of a triggered alert condition or correlated group of incidents. Issues have a severity, an owner, and a lifecycle from created to closed, and can route to destinations like email, Slack, or PagerDuty.

A logical grouping of entities that should be monitored and alerted on together, such as all services behind one product. Workloads aggregate health scores and let you apply alert policies across many entities at once.

A key-value label applied to entities for organization and filtering, such as team or environment. Tags drive dashboard filters, workload membership, and which alert policies apply to which entities.

A host-level agent that collects CPU, memory, disk, network, and process metrics from servers, containers, and cloud instances. It reports into the Infrastructure event types and supports on-host integrations for services like databases and web servers.

A plugin configuration that runs alongside the infrastructure agent to collect metrics from third-party software such as MySQL, NGINX, or Redis. Each integration reports its own event types and sample dashboards.

A timestamped text record forwarded to New Relic, either through an agent, the infrastructure agent's log forwarding, or the Log API. Logs are stored in the Log event type and can be parsed and linked to traces.

A configuration that extracts structured attributes from raw log lines using patterns such as Grok or regex. Parsed fields become queryable attributes, enabling filtering and faceting that plain text does not allow.

A numeric time-series measurement such as CPU percentage or request count. Metrics are aggregated over time intervals, are cheaper to store than raw events, and are stored in dimensional metric event types queryable with NRQL.

The period for which each telemetry type is stored before aging out, varying by data type and account tier. Retention affects how far back NRQL queries and dashboards can look.

The practice of storing only a fraction of high-volume telemetry, such as transaction traces or spans, to control cost and volume. Sampled data is statistically representative but may miss individual specific requests.