ECS 2027
31/05/2027, Congress Center Düsseldorf
Düsseldorf, Germany
Same service, two clouds: OpenTelemetry observability for .NET and Node backends on Azure and AWS
Teams that inherit or choose a mixed estate often end up with fragmented tooling: Application Insights on one side, CloudWatch on the other, and no way to follow a request that crosses both. OpenTelemetry offers a vendor-neutral layer, but the real work is in the details.
I'll walk through instrumenting a .NET service and a Node.js/TypeScript service with OpenTelemetry, then exporting to Azure Monitor and to AWS (via the AWS Distro for OpenTelemetry, CloudWatch and X-Ray). We'll look at context propagation across service and cloud boundaries, sensible sampling for high-traffic systems, and how to keep cost and cardinality under control. Then we'll define SLOs and alerts, and use them to debug a realistic latency regression end to end. Finally, I'll cover the honest trade-offs: what's lost by going neutral, and when the native tooling is still the better choice.
Learning objectives:
- Instrument .NET and Node services with OpenTelemetry once
- Export and correlate traces, metrics and logs in Azure Monitor and AWS
- Propagate trace context across clouds and asynchronous boundaries
- Apply sampling and cardinality controls for high-traffic workloads
- Choose when to stay native and when to stay neutral