Metrics vs Events: A Conversation About Controlling Volume

The discussion between Steven E. Harris and Ben Hartshorne in Honeycomb's Community Slack explores the differences between Honeycomb's event model and the traditional metrics instrumentation model, highlighting the former's approach to handling high cardinality and data volume through aggregation and sampling. While traditional metrics models, often using aggregation, maintain fixed storage costs by periodically publishing data, Honeycomb's event-based system offers more detailed insights but may lead to higher storage costs without proper sampling. Ben emphasizes that while Honeycomb can replace a significant portion of traditional metrics use, some scenarios still benefit from metrics, particularly where high throughput and low differentiation are involved. The conversation underscores that both models have their strengths, and integrating them can provide comprehensive insights into system behaviors, suggesting that Honeycomb's approach can complement rather than entirely replace traditional metrics systems.