Telemetry

What it is

Telemetry is evidence a system emits about what it is doing: numeric measurements, log events, request traces, dependency calls, or usage signals. It is the raw material for monitoring and diagnosis, not the diagnosis itself.

Why it exists

Without telemetry, a team is forced to guess why a service is slow or failing. With useful telemetry, it can measure a symptom, find affected requests, correlate events, alert the right people, and test whether a fix worked.

Key ideas

• Kinds of telemetry:
  • Metrics — numeric, time-series measurements (CPU, latency, request count).
  • Logs — text/diagnostic events, often queryable with KQL.
  • Traces — end-to-end request/dependency paths (distributed tracing).
  • Usage/analytics telemetry — how users interact (page views, features used).
• Sources differ. Azure resources emit platform data through Azure Monitor; application code can emit application telemetry through Application Insights or an instrumentation library.
• A path is needed. Telemetry must be emitted, collected, stored, queried or visualized, and acted on. Not every data source is enabled or routed automatically.

Northwind example

Northwind’s checkout app records request duration, failed dependency calls, and a correlation ID. Azure Monitor supplies platform metrics for the hosting service. When checkout is slow, the team first sees a latency increase on a metric chart, then uses traces and logs tied to the correlation ID to find whether the payment dependency or the app is failing. Those signals serve different roles but describe the same customer journey.

Exam notes

• Metrics, logs, traces, and events answer different questions; choose the signal before choosing the chart or alert.
• Platform telemetry and application telemetry can be correlated, but one does not replace the other.
• Telemetry becomes useful when it supports a dashboard, alert, or feedback loop.

Related

continuous-feedback · azure-monitor · application-insights · azure-monitor-metrics · log-analytics · alerting · feedback-loop

📘 Source: Microsoft Learn — Telemetry