Managed tail-based sampling with RaaS
Refinery as a Service (RaaS) is a managed service that provides Honeycomb's trace-aware, tail-based sampling engine without the infrastructure overhead. We deploy, operate, and observe a dedicated Refinery instance for you that keeps the traces that matter while cutting the noise.
Why you'll love RaaS
Keep the traces that matter
Tail-based sampling examines complete traces before deciding what to keep, so errors, slow requests, outliers, and novel patterns always make it to Honeycomb. Head-based sampling decides before the request finishes and can sample away the traces you need most.
Control observability spend
Reduce event volume before it reaches Honeycomb while preserving full fidelity on the moments you care about. Volume-aware sampling keeps your costs predictable, even when traffic isn't.
Zero Refinery operations
No capacity planning, no upgrades, no on-call rotation for sampling infrastructure. Honeycomb's Field Reliability and Support Engineering teams run your dedicated Refinery instance 24/7 and you get full visibility into every sampling decision via metadata on your telemetry.
How it works
Three steps from your pipeline to high-value data in Honeycomb.
Send traces via OpenTelemetry
Your apps export traces through your existing OpenTelemetry pipeline. Use it to redact sensitive fields before egress.
Refinery samples intelligently
Your dedicated Refinery instance buffers complete traces, keeping errors, slow requests, and other interesting patterns.
High-value data in Honeycomb
Only the traces that matter reach Honeycomb, reducing volume while preserving fidelity.
Why choose Refinery as a Service?
Sampling that understands traces
Refinery combines dynamic, rules-based, and throughput-based sampling to help you keep 100% of errors, sample healthy 2xx traffic at 1-in-1,000, and cap total span throughput based on the full contents of each trace.
Single-tenant by design
Your Refinery runs in a dedicated namespace on isolated pods with no shared resources. With AWS PrivateLink, traffic stays on the AWS backbone and never touches the public internet. SOC 2 Type II compliant infrastructure.
Operated by the people who built it
Honeycomb's Field Reliability and Support Engineering teams deploy, scale, upgrade, and monitor your Refinery instance with 24/7 support.
Full visibility into every decision
RaaS isn't a black box. Sampling decisions are observable via Honeycomb telemetry, so you can see exactly what's kept and what's dropped.
The Engineer’s Guide to Sampling in the Age of AI
Without proper management, the sheer volumes of logs, metrics, and traces can overwhelm storage and processing systems. By implementing strategies like sampling, engineers can optimize their telemetry data by retaining the most valuable information while reducing noise and costs. In this guide, you'll learn about:
- What sampling is
- Sampling methods you can leverage
- When you need sampling and when you can avoid it
- Configuring your sampling tool
- How to get started with sampling
Want to know more?
Talk to our team to arrange a custom demo or for help finding the right plan.