Telemetry Optimization Lab

A working exploration of the question most large observability environments eventually run into: not "can we collect this telemetry," but "should we, at this cardinality, at this volume, at this cost."

Planned scope

• Metric cardinality — reproducing common cardinality explosions (unbounded labels, per-request IDs) and measuring their cost and query impact directly.
• Noisy logs — identifying and suppressing high-volume, low-value log lines without losing the signal needed during an incident.
• Trace sampling — comparing head-based and tail-based sampling strategies and their effect on both cost and the ability to catch rare, expensive traces.
• Unnecessary telemetry — a practical audit method for finding metrics, logs, and traces nobody queries.

Why this project

This is the day-to-day reality of sizing and optimization work in observability POVs — customers rarely have a cardinality or cost problem in the abstract, they have a specific noisy job or an unbounded label that's driving their bill. This lab is meant to make those patterns concrete and repeatable, rather than theoretical.

Status

Planning stage. Findings and reproducible examples will be added here as the lab is built out.