Virtual EnvoyCon 10 Year Anniversary Edition
Oct 14, 2:00 – 4:00 PM (UTC)
Virtual Project Events (Hosted by CNCF)
Join us at Virtual EnvoyCon 10 Year Anniversary Edition, the go-to event for Envoy practitioners of all experience levels. Whether you're just getting started, refining your expertise, or pushing the boundaries of what's possible, this community-driven virtual conference offers deep technical content and practical insights from both contributors and users.
This event is all about real-world experience—no product pitches, just meaningful conversations, collaboration, and innovation within the Envoy open-source ecosystem.
CFP Closes: Wednesday, July 22
Speaker Notifications: Monday, August 3
Schedule Announced: Wednesday, August 5
Discover Envoy Proxy’s powerful capabilities for high-performance proxying and extensibility.
Explore Envoy Gateway, a Kubernetes-native control plane that simplifies working with Envoy Proxy.
Learn how Envoy is evolving to handle modern AI traffic and what’s next for the community.
A CNCF graduated project since 2017, Envoy powers diverse deployment models including edge proxies, service mesh sidecars, and load balancing.
Join us for a day of engaging talks and case studies, and connect directly with maintainers, community leaders, and adopters. Let’s build the future of Envoy—together!
October 14, 2026 · 10:00 AM – 1:00 PM EST Virtual · Free · 6 talks · 8 speakers · Single track
Hosts: Erica Hughberg (Tetrate) and Rohit Agrawal (Databricks)
Andrew Katsikas — Senior Software Engineer, Pelotech Configuring Envoy · Intermediate
Ingress NGINX is feature-frozen and unpatched, and Gateway API is the clear next step. Most migration guides stop once traffic moves, leaving the hard part unsolved: not dropping requests.
A case study of a real customer migration on AWS — how Gateway API controllers were evaluated against CNCF alignment, mTLS and request buffering, why the first cutover still had a DNS-driven downtime window, and how weighted DNS via ExternalDNS and Route 53 let Ingress and HTTPRoute run side by side for a zero-failed-request swap with a trivial rollback path.
Closes with the multi-namespace hostname problem their own testing missed, and how Gateway API 1.5's ListenerSet — already supported in Envoy Gateway — separates infrastructure and application concerns.
Basundhara Chakrabarty — Software Engineer, Databricks Rohit Agrawal — Sr. Staff Software Engineer, Databricks Extending Envoy · Beginner
Dynamic Modules let you extend Envoy in Rust without maintaining a fork or writing C++. Most examples out there are HTTP filters, but the extension surface is much wider — stats sinks, access loggers, health checkers and more, often small changes that pay off quickly.
Two examples built at Databricks: a stats sink that moves Prometheus metric collection off Envoy's main thread, so proxies running thousands of clusters stop starving config updates on every scrape; and a logger that replaces three separate logging paths with one Rust module handling rule evaluation, sinks, sampling and log-derived metrics on a Tokio runtime instead of the request path.
The pattern is the same both times: find work stuck on a hot path and move it somewhere it can scale on its own.
Lightning talk
Tianyu Xia — Software Engineer, Google Envoy and AI Traffic · Intermediate
LLMs and AI agents are shifting the enterprise networking paradigm. Real-time streaming, massive payloads and evolving protocols — the OpenAI API, Model Context Protocol, Agent-to-Agent communication — push traditional HTTP-centric proxies to their limits. Forcing these workloads into Envoy's existing HTTP filter chain produces redundant payload parsing, high CPU and memory overhead, and fragmented extensions.
The architecture behind Native AI Support in Envoy (issue #44681): a high-performance centralized AI protocol parser covering LLM model transformations (Gemini, OpenAI, Anthropic), MCP/A2A authentication, transcoding and inference gateway capabilities — treating AI protocols as a distinct, native layer rather than an L7 HTTP afterthought.
Xin LI — Staff Software Engineer, Traffic Team, LinkedIn Ruei-Bang Chen — Senior Software Engineer, Traffic Team, LinkedIn Envoy Project Fundamentals · Intermediate
Starting from xDS configuration, LinkedIn built a service discovery and routing platform on Envoy that powers the edge proxy connecting external traffic from the LinkedIn website and mobile apps to internal services.
Architecture, key challenges, solutions, and the current state of the platform.
Anurag Aggarwal — Senior Software Engineer, Tetrate Envoy at Scale · Intermediate
A production story of scaling a multi-tenant AI gateway on Envoy to thousands of tenants — and the memory and config-size walls that stop you long before you expect.
Give every tenant their own models, limits and providers and config grows as tenants × models. Two walls hit early: the proxy's memory footprint, since each upstream cluster costs heap; and a low ceiling on how much routing config a gateway can hold, breached by only a handful of tenants.
How both were diagnosed, the stopgap that kept production alive, and the fixes — shared clusters, per-tenant policy moved onto the request, and changes contributed upstream to Envoy Gateway.
Lightning talk
Hong Sng — Software Engineer, Databricks Extending Envoy · Intermediate
Databricks manages millions of virtual machines every day, supporting data (Spark, storage, streaming) and AI (apps, agents, model serving) workloads for over 60% of the Fortune 500.
Envoy is extended with Dynamic Modules to provide performant access controls for governance.
Tetrate
Envoy AI Gateway Maintainer
Databricks
Software Engineer
CONTACT US
