SE Radio 717: Eric Tschetter on Decoupling Observability

Observability technology is transitioning from monolithic, vendor-locked stacks toward a decoupled three-layer architecture consisting of data ingestion, storage, and query-visualization layers. This evolution mirrors the historical development of business intelligence, where separating the data platform from the interaction layer fosters greater flexibility and prevents vendor lock-in. Eric Schetter, Chief Architect at Imply and co-author of Apache Druid, argues that standardization should focus on query languages—such as SQL, PromQL, and LogQL—rather than attempting to enforce a universal data schema, which remains impractical due to the inherent variability of log data. By leveraging cloud object stores for cost-effective storage and enabling independent compute for specific query patterns, organizations can maintain interoperability across diverse tools, allowing teams to retain preferred workflows while gaining access to broader, unified data sets.