Airbnb Switches to OpenTelemetry, Handles 100M Metrics per Second

Why It Matters

Airbnb’s migration demonstrates that open‑source observability tools can meet the most demanding data ingestion requirements, challenging the dominance of proprietary APM vendors. By standardizing on OpenTelemetry, Airbnb reduces operational complexity, gains flexibility to switch downstream storage solutions, and opens the door to richer, cross‑service insights. The move also validates the scalability of VictoriaMetrics’ vmagent, encouraging other enterprises to consider it for high‑throughput metric workloads. The broader market impact includes heightened pressure on commercial monitoring platforms to lower prices and improve interoperability. As more firms adopt OpenTelemetry, a unified telemetry data format could emerge, simplifying integration across cloud providers, SaaS tools, and on‑premise systems. This convergence may accelerate the development of advanced analytics, AI‑driven anomaly detection, and automated remediation capabilities that rely on consistent, high‑resolution observability data.

Key Takeaways

• Airbnb migrated from StatsD and Veneur to an OpenTelemetry‑based pipeline.
• The new system ingests over 100 million metric samples per second in production.
• Components include the OpenTelemetry Protocol, OpenTelemetry Collector, and VictoriaMetrics’ vmagent.
• Migration required rewriting instrumentation and adjusting alerting across multiple teams.
• Airbnb plans to extend the stack to traces and logs for a unified observability platform.

Pulse Analysis

Airbnb’s decision to overhaul its metrics pipeline reflects a maturation of the open‑source observability stack that began with the OpenTelemetry project in 2019. Early adopters used the protocol for low‑volume workloads; today, the platform can sustain data rates that were once the exclusive domain of enterprise‑grade APM solutions. This evolution reduces the cost barrier for large‑scale telemetry, as open‑source components eliminate licensing fees and enable tighter integration with custom data pipelines.

Historically, companies like Airbnb built proprietary aggregation layers to meet unique performance needs, but those solutions often became technical debt as the ecosystem evolved. By moving to OpenTelemetry, Airbnb not only gains access to a vibrant community of contributors but also aligns its observability strategy with industry standards, simplifying talent acquisition and cross‑team collaboration. The choice of VictoriaMetrics’ vmagent is particularly noteworthy; its ability to handle 100 M samples per second validates the performance claims of newer open‑source time‑series databases, potentially reshaping the competitive landscape against incumbents such as Datadog and Splunk.

Looking forward, the real test will be how Airbnb integrates traces and logs into the same pipeline while preserving the low‑latency characteristics required for real‑time monitoring. If successful, the unified stack could become a blueprint for other high‑traffic platforms seeking to consolidate observability under a single, open framework. The ripple effect may drive broader adoption of OpenTelemetry, spur innovation in downstream analytics, and accelerate the shift toward more transparent, interoperable monitoring ecosystems across the Big Data industry.