Self-Hosted Observability: The Migration Every Team Is Doing in 2026
Teams are migrating from expensive SaaS observability tools like Datadog to a self-hosted stack of OpenTelemetry Collector, ClickHouse, and Grafana to monitor their LLM-powered applications.
Why it matters
This migration represents a significant cost-saving opportunity for teams deploying LLM-powered applications, allowing them to better control their observability costs.
Key Points
- 1Datadog's pricing model is not suitable for LLM applications that generate an order of magnitude more telemetry data
- 2The 2026 self-hosted stack includes OpenTelemetry Collector, ClickHouse, and Grafana, with an optional Langfuse container
- 3OpenTelemetry's GenAI semantic conventions enable easy migration by standardizing the data format
Details
LLM features generate significantly more observability data than traditional HTTP services, leading to rapidly growing Datadog bills for teams. The self-hosted stack of OpenTelemetry Collector, ClickHouse, and Grafana provides a cost-effective alternative that can ingest, store, and visualize the high volume of LLM-related telemetry data. The recent acquisition of Langfuse by ClickHouse further integrates the trace exploration capabilities into the stack. The standardization of OpenTelemetry's GenAI semantic conventions enables easy migration, as the data format is now widely adopted by LLM platforms and observability tools.
No comments yet
Be the first to comment