Summary
OpenTelemetry (OTel) is a vendor-neutral, CNCF-graduated standard for generating and collecting telemetry data—traces, metrics, and logs. It provides a common set of APIs, SDKs, and a wire protocol (OTLP) so that instrumentation is written once and the resulting data can be exported to any compatible backend.
What is OpenTelemetry?
OpenTelemetry standardises how applications emit observability data. Code is instrumented with the OTel SDK, which produces telemetry in a uniform format; an optional OTel Collector then receives, processes, and routes that data to one or more backends such as Grafana Tempo, Jaeger, Honeycomb, or Dynatrace. Because everything speaks the OpenTelemetry Protocol (OTLP), switching or adding a backend requires no change to application code.
OTel also implements the W3C Trace Context standard for propagating a trace ID across service boundaries, which is what makes end-to-end distributed tracing possible in a microservices architecture. More recently it added semantic conventions for AI and GenAI workloads, giving teams a standard way to capture spans and token usage from LLM-based applications and agents.
OpenTelemetry reached CNCF "graduated" status in 2026, formalising its position as the de-facto observability standard.
Why is OpenTelemetry relevant?
- No vendor lock-in: Instrument once, then send the same data to any OTLP-compatible backend
- End-to-end tracing: W3C Trace Context propagation links requests across services for true distributed tracing
- AI observability: GenAI semantic conventions standardise how agent and LLM activity is captured
- Audit and compliance: Traces provide a natural, standardised audit trail for regulated environments