I’m wondering if a similar approach could work for observability data — especially logs. Hot storage is expensive, and much of the data may not be queried after a short period. Moving unused logs to warm or cold storage (or dropping them) could potentially save a lot.

Has anyone tried this kind of tiering or aging strategy for logs or metrics? Would love to hear how you approached it — tools, heuristics, lessons learned. Thoughts and speculation are also very welcome!