OpenTelemetry: Skyscanner's Path to Future-Proof Observability

Decoupling Observability for Engineering Excellence

The presentation effectively articulates the challenges of traditional observability approaches and positions OpenTelemetry as a solution for decoupling instrumentation from vendors, thereby reducing technical debt and improving engineer autonomy. The emphasis on treating the observability platform as a product with engineers as customers is a crucial cultural shift. The architectural shift towards correlated signals, rather than siloed pillars (metrics, traces, logs), powered by OpenTelemetry's semantic conventions and standardized APIs/SDKs, is well-explained. The concept of "minimal viable telemetry" baked into internal libraries or base images is a practical approach to ensuring consistent observability across a large organization.

The discussion on "platform boundary" and pushing the platform's influence closer to the application side via SDK configuration is particularly insightful. This allows for greater consistency and adherence to standards without sacrificing developer flexibility. The mention of reusable modules (Terraform, scripts, templates) for rolling out standards like SLO configurations and dashboards highlights a mature approach to platform engineering. The "agentless" vision for telemetry, where it's baked into libraries, is also a significant advantage for reducing overhead and simplifying deployment.

While the presentation highlights the benefits of OpenTelemetry, a deeper dive into the challenges of migrating existing instrumentation or dealing with particularly complex legacy systems might have added further value. The practicalities of maintaining and evolving the OpenTelemetry SDK configurations across a vast microservices landscape, especially concerning versioning and compatibility, could also be explored more thoroughly. Nonetheless, the overall message is clear: OpenTelemetry offers a robust, vendor-agnostic foundation for building scalable and future-proof observability, directly empowering engineering teams.

Key Points

• Moving to OpenTelemetry decouples instrumentation from vendors, reducing technical debt and context switching for engineers.
• Treating the observability platform as a product with engineers as customers is key to reducing incident rates and improving developer experience.
• OpenTelemetry enables correlated signals (metrics, traces, logs) instead of siloed pillars, providing richer context for debugging.
• Standardized APIs, SDKs, and semantic conventions in OpenTelemetry lower cognitive load for engineers and ensure consistency.
• The "platform boundary" can be pushed closer to the application via SDK configuration, enabling "minimal viable telemetry" through shared configurations or internal libraries.
• Reusable modules (Terraform, scripts) can standardize SLOs, dashboards, and alert configurations, facilitating frictionless adoption of best practices.

---

📖 Source: Presentation: Building a Future-Proof Observability Platform to Empower Engineers