OpenTelemetry Architecture and Components
What you'll learn: How OpenTelemetry's modular architecture connects APIs, SDKs, collectors, and exporters to create a unified observability pipeline.
The Big Picture
OpenTelemetry (OTel) isn't a monolithic tool—it's an ecosystem of cooperating components that work together to capture, process, and export telemetry data (traces, metrics, and logs). Think of it like a factory assembly line: raw data enters at one end, gets shaped and enriched along the way, then exits to your chosen backend.
Core Components
API
The API defines what telemetry operations look like—creating spans, adding attributes, recording metrics. It's a specification that application code interacts with. This abstraction means you can swap implementations without changing your code.
SDK
The SDK is the implementation of the API. It actually creates spans, manages trace context, and handles sampling decisions. The SDK is what you configure and customize for your environment.
Collectors
The Collector is a standalone service that receives telemetry from applications, processes it (filtering, batching, enriching), and forwards it to backends. It decouples your applications from backend-specific logic. You can run collectors as agents (per-host) or as centralized gateways.
Exporters
Exporters are pluggable modules that send telemetry to specific backends—Jaeger, Zipkin, Prometheus, cloud vendors, etc. They translate OTel's internal format into what each backend expects.
How They Work Together
Your application uses the API → implemented by the SDK → sends data to a Collector → which uses Exporters to route data to backends. This separation allows you to instrument once and switch backends without re-instrumenting.
Key Takeaway: OpenTelemetry's modular architecture (API, SDK, Collector, Exporters) creates vendor-neutral instrumentation where applications produce standardized telemetry that can be routed anywhere through configuration, not code changes.