Telemetry types

Before we dive into specific data sources, let’s talk about types of data, because different data sources are optimized for different things.

Metrics are numbers over time: CPU at 80%, 500 requests per second, 99th percentile latency of 200 milliseconds. Logs are event records: error messages, access logs, audit trails.

Traces show how a request flows through your system. A user clicks a button, the request hits the API, the API calls the database, and the response returns. Profiles show what your code is actually doing, which functions are slow, and where memory is being allocated.

Most teams start with metrics and logs. Those solve the majority of observability problems. Traces and profiles typically come later as your observability practice matures.

The key point: your data type determines which data sources make sense. Prometheus is built for metrics. Loki is built for logs. Choose the right tool for your data.