On-Prem Load Balancing Reimagined: Serving 20 Million QPS With Gateway API and EnvoyGateway#
KubeCon + CloudNativeCon North America 2025 digest Program Core pick 10–13 November 2025
Isaac Wilson, The Trade Desk — Core
- Why: One of the few talks with a real scale number (20M QPS) and an explicit "things we broke" framing — a bare-metal HAProxy to Kubernetes/Envoy migration at the high end of on-prem load balancing.
- What: How an established bare-metal HAProxy Community Edition load-balancing tier was rebuilt as an Envoy Gateway / Gateway API platform, including upstream Envoy contributions. (from abstract only)
- How: Migrated core services and refactored service discovery; extended Envoy Gateway with circuit breakers and zone-aware routing for topology-aware traffic distribution and resilience, contributing some of that work upstream. (from abstract only)
- Where applicable: On-prem / bare-metal load balancing at large scale; broadly applicable to Gateway API + Envoy Gateway adopters, though the 20M QPS economics are specific to The Trade Desk's ad-tech workload.
