The OpenTelemetry community has launched OpenTelemetry Profiles as a public Alpha signal, establishing the first industry-wide standard for continuous production profiling. This fills a long-standing gap, providing a unified, vendor-neutral framework alongside the existing standards for traces, metrics, and logs. The goal is to make low-overhead performance analysis a routine part of observability, helping teams troubleshoot incidents and optimize software efficiency directly in production environments.

A core achievement is the new OTLP Profiles data format, designed for efficiency and compatibility. It features deduplicated stack traces and dictionary tables to minimize size, while maintaining interoperability with the popular pprof format via a new native translator. For immediate use, an eBPF-based profiling agent—donated by Elastic—now integrates with the OpenTelemetry Collector. This agent supports major runtimes like Go, Node.js (ARM64), .NET 9/10, and BEAM (Erlang/Elixir) without requiring code instrumentation.

Profiles are being woven into the broader OpenTelemetry ecosystem. The Collector can now receive and augment profile data with Kubernetes metadata, and the resource model is updated for efficient cross-signal correlation with logs and traces. While production-ready backends are still emerging from vendors, Elastic's experimental devfiler desktop app offers a way to experiment now. This Alpha release marks a concrete step toward making continuous profiling a standard, integrated practice rather than a niche toolset.