Retroactive Sampling with OpenTelemetry: Cut 90% Distributed Traci... Roman Khavronenko & Zhu Jiekun

Roman Khavronenko and Zhu Jiekun share a retroactive sampling approach in OpenTelemetry to reduce distributed tracing bandwidth usage.

Don't miss out! Join us at our next KubeCon + CloudNativeCon events in Mumbai, India (18-19 June, 2026), Yokohama, Japan (29-30 July, 2026), and Shanghai, China (8-9 September, 2026). Connect with our current graduated, incubating, and sandbox projects as the community gathers to further the education and advancement of cloud native computing. Learn more at kubecon.io Retroactive Sampling with OpenTelemetry: Cut 90% Distributed Tracing Bandwidth Usage - Roman Khavronenko & Zhu Jiekun, VictoriaMetrics Have you considered the traffic consumed by tail-sampling traces in large-scale setups, where: 1. Apps run across clusters/AZs. 2. Trace spans from apps must be sent to an intermediate OpenTelemetry collector for tail-sampling. While cross-cluster/AZ traffic is unavoidable, most spans sent over the network get discarded, wasting much bandwidth. In this session, we’ll share the retroactive sampling approach: each OpenTelemetry collector buffers spans from the node/cluster it resides in, sending only necessary data to the centralized trace backend for decision-making. The backend then requests raw spans only if the trace is marked as sampled. By joining our session, you'll have more insights into: 1. How tail-sampling burns your costs and how retroactive sampling cuts them by 9x. 2. How to use the OpenTelemetry collector for retroactive sampling: architecture design and benefit-maximizing strategies. 3. Practical ways to adopt it in your existing architecture, plus trade-off analysis.