NEW YORK — What happened: Industry feedback compiled from DevOps leadership forums indicates a growing backlash against indiscriminate data collection under the OpenTelemetry standard.
Why it matters: Observability was supposed to solve system blindness by collecting every metric, trace, and log available. Instead, it has created a new operational burden: filtering through petabytes of noise to find the handful of signals that actually indicate a customer-facing outage.
Deep dive: The open-source OpenTelemetry project has successfully standardized how telemetry data is collected and exported. However, many engineering teams implemented the 'collect everything now, figure it out later' philosophy. This has resulted in massive telemetry ingestion pipelines that consume significant compute resources just to index logs that are rarely read.
Report check: Reports suggest that telemetry storage now accounts for nearly 20% of total internal infrastructure budgets at mid-sized SaaS companies. Industry surveys confirm widespread adoption of OpenTelemetry, but also highlight a shift toward aggressive client-side sampling and dynamic head-based filtering to reduce volume.
Open questions: How will tooling vendors adapt their pricing models if enterprises successfully reduce their raw log volume by 70% through better telemetry filtering?