New Relic Performance Engineer flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Apdex formula

Apdex = (satisfied + tolerating/2) / total requests, where satisfied requests are at or below the threshold T, tolerating requests fall between T and 4T, and frustrated requests exceed 4T.

Transaction trace purpose

A transaction trace records a single request's segment breakdown, showing time spent in database calls, external services, and application code to locate the dominant latency contributor.

Distributed tracing headers

Distributed tracing links segments across services using trace and span identifiers propagated in request headers, including W3C traceparent, so a request's full path is reconstructable.

Error group fingerprinting

Error groups cluster similar errors using the error class and a stack trace fingerprint rather than the raw message, so variable text in messages does not fragment grouping.

Key transaction

A key transaction is a designated business-critical transaction tracked separately from the application aggregate, with its own alerting, dashboards, and Apdex score.

APM 360 scope

APM 360 combines application performance monitoring with infrastructure, browser, mobile, synthetic, and applied intelligence data, correlated through the entity model.

SLO error budget

An SLO defines a target ratio of good events to total events over a rolling window; the error budget is the allowed bad-event fraction, and remaining budget tracks consumption.

Thread profiler use

The thread profiler samples stack traces of running threads to attribute CPU time to specific methods, useful when slowness is spread across many requests rather than one trace.

Core Web Vitals thresholds

At the 75th percentile, LCP should be 2.5 seconds or less, CLS 0.1 or less, and INP 200 milliseconds or less; these are the standard good-experience thresholds.

Synthetic monitor types

Synthetic monitors include ping, simple browser, and scripted browser checks that run on schedules from selected locations to verify availability and step-level user flows.

Private synthetic locations

Private locations are agent containers you host yourself, used to monitor internal or firewall-protected endpoints and to control the source geography of synthetic checks.

Session trace vs replay

A session trace reconstructs a user session's page loads, AJAX calls, and JS errors as data; session replay adds a visual reconstruction of the recorded interaction.

JS error source maps

Source maps uploaded to New Relic de-minify minified JavaScript bundle stack traces so browser JS errors display original file names, line numbers, and function names.

Browser-backend correlation

Browser instrumentation propagates distributed tracing headers to the backend, letting the browser app entity link to the APM application and connect front-end to server-side latency.

Mobile crash symbolication

Mobile crash reports arrive with memory addresses; symbolication using iOS dSYM files or Android mapping files resolves them into readable class and method names.

TTFB and FCP

Time to first byte measures server response latency before the browser receives content; first contentful paint marks when the first text or image is rendered.

Anomaly alert conditions

Anomaly conditions learn a seasonal baseline for a signal and open incidents when the observed value deviates beyond the learned band, rather than against a fixed threshold.

Alert severity levels

Alert conditions produce incidents at Critical, Warning, or Info severity; incidents are grouped by policy and condition and close automatically when the condition clears.

NRQL alert evaluation offset

The evaluation offset shifts the query window back in time so it covers only complete data, and it must exceed the aggregation window to avoid evaluating partial buckets.

Muting rules

Muting rules suppress notifications for a defined scope and schedule without deleting the alert condition, and can be configured to also suppress incident creation.

Workload health rollup

A workload groups entities for prioritized alerting, and its health reflects the worst status among member entities unless custom weighting changes the rollup.

Anomaly learning period

Anomaly detection requires a learning period, typically several days, before baselines stabilize; conditions on newly created signals may not fire reliably until then.

Notification destinations

Destinations route incidents to channels such as email, webhooks, Slack, PagerDuty, and ServiceNow; workflows bind a destination to filters on incident attributes.

Alert condition query shape

An NRQL alert condition query must return a single scalar value per evaluation, which is then compared against the configured threshold to decide whether to open an incident.

SystemSample event

The infrastructure agent reports SystemSample events containing CPU, memory, disk, and network metrics for a host, queried with NRQL using FROM SystemSample.

Dimensional metric syntax

Dimensional metrics are written as metricName{tag1: value} and queried with FROM Metric, distinguishing them from event-based infrastructure data such as SystemSample.

Kubernetes integration events

The nri-kubernetes integration runs as a DaemonSet and collects K8sNodeSample, K8sPodSample, and K8sContainerSample events for cluster, pod, and container monitoring.

Cloud integration latency

Cloud provider integrations poll provider APIs and ingest metrics at provider granularity, often one to five minutes, so they are not real-time and lag behind agent data.

Host golden signals

Host golden signals are CPU utilization, memory used percentage, disk I/O wait, network throughput, and filesystem free space; alert on sustained deviation, not single samples.

Log event type

Logs land in the Log event type after ingestion through the infrastructure agent, Log API, or forwarders, and are queried with NRQL using FROM Log.

Log parsing rules

Log parsing rules transform raw log messages into structured attributes at ingest time, before storage, so unparsed fields remain searchable only as the raw message.

Log-to-metric rules

Log-to-metric and log-to-event rules derive metrics or events from log patterns, enabling alerting on log content without storing high-cardinality log data as metrics.