Phase 6 — References

The whole reading list for the Cloud Gateway role, grouped by theme. Per-lab references.md files go deeper; this is the master index.

The Netflix talks & posts named in the JD (read these first)

• Curbing Connection Churn in Zuul — Netflix TechBlog, 2023. The per-event-loop connection pool + subsetting work; the headline result (≈8× fewer TCP opens, churn from thousands/s to ~60/s). Directly named in the JD. https://netflixtechblog.com/curbing-connection-churn-in-zuul-2feb273a3598 — Arthur Gonigberg's companion write-up: https://arthur.gonigberg.com/2023/10/03/curbing-connection-churn/
• Pushy to the Limit: Evolving Netflix's WebSocket proxy for the future — Netflix TechBlog, 2024. Hundreds of millions of concurrent WebSocket connections, push registry on KeyValue (was Dynomite), Kafka message processor, 60k→200k→400k conns/node. https://netflixtechblog.com/pushy-to-the-limit-evolving-netflixs-websocket-proxy-for-the-future-b468bc0ff658
• Scaling Push Messaging for Millions of Devices — InfoQ / re:Invent 2018. The original Zuul Push / Pushy architecture. https://www.infoq.com/news/2018/07/zuul-push-messaging/
• Zuul 2: The Netflix Journey to Asynchronous, Non-Blocking Systems — Netflix TechBlog. The Netty rewrite; event loop vs thread-per-request; ~25% throughput gain; 80+ Zuul clusters, >1M rps. https://netflixtechblog.com/zuul-2-the-netflix-journey-to-asynchronous-non-blocking-systems-45947377fb5c
• Zero Configuration Service Mesh with On-Demand Cluster Discovery — Netflix TechBlog. Envoy + a control plane that discovers clusters lazily; the service-mesh direction of the Studios-security talk. https://netflixtechblog.com/zero-configuration-service-mesh-with-on-demand-cluster-discovery-ac6483b52a51
• Titus, the Netflix container management platform — TechBlog + ACM Queue "Titus: Introducing Containers to the Netflix Cloud." The Mesos→Kubernetes story behind "Managing Netflix's Compute with K8s." https://netflixtechblog.com/titus-the-netflix-container-management-platform-is-now-open-source-f868c9fb5436 · https://queue.acm.org/detail.cfm?id=3158370
• Container Runtime Customization at Netflix (NRI & OCI Hooks) — KubeCon NA. containerd NRI + OCI hooks for per-workload runtime customization while staying K8s-compatible. https://github.com/containerd/nri

Source code worth reading

• Netflix/zuul (Java/Netty) — the reference open-source API gateway. Read BaseZuulFilter, the filter types (inbound/endpoint/outbound), and the Netty ChannelHandler pipeline. https://github.com/Netflix/zuul
• envoyproxy/envoy (C++) — the canonical L4/L7 proxy. Read the listener → filter chain → cluster model; one HTTP filter end to end. https://github.com/envoyproxy/envoy
• envoyproxy/go-control-plane (Go) — build an xDS server (gw-08). https://github.com/envoyproxy/go-control-plane
• kubernetes-sigs/gateway-api + controller-runtime (Go) — the Gateway API CRDs and the operator framework (gw-10). https://github.com/kubernetes-sigs/gateway-api · https://github.com/kubernetes-sigs/controller-runtime
• containernetworking/cni + cilium/cilium — the CNI spec and an eBPF dataplane (gw-09). https://github.com/containernetworking/cni
• spiffe/spire — workload identity / SVID issuance (gw-07). https://github.com/spiffe/spire

Protocol specs (the L4/L7 canon)

• TCP — RFC 9293 (consolidated). UDP — RFC 768.
• HTTP/1.1 — RFC 9110 (semantics) + RFC 9112 (syntax).
• HTTP/2 — RFC 9113. HPACK — RFC 7541.
• HTTP/3 — RFC 9114. QUIC — RFC 9000. QPACK — RFC 9204.
• WebSocket — RFC 6455.
• TLS 1.3 — RFC 8446. PROXY protocol — HAProxy spec v1/v2.
• gRPC — the grpc-over-http2 wire spec in grpc/grpc.
• W3C Trace Context — https://www.w3.org/TR/trace-context/

Books & long-form

• Site Reliability Engineering (Google) — load shedding, retry budgets, "Addressing Cascading Failures," handling overload.
• Marc Brooker's blog — timeouts, retries, jitter, metastable failures. https://brooker.co.za/blog/
• Systems Performance (Brendan Gregg) — USE method, the kernel network stack, eBPF.
• Kubernetes Networking docs + Networking and Kubernetes (O'Reilly).
• Envoy docs — the best free explanation of a modern data plane: https://www.envoyproxy.io/docs

Cross-phase links

• db-16…db-20 (consensus) — the control plane's source of truth is consensus-backed; xDS config propagation is a distributed-consistency problem in disguise. gw-08 references db-17 directly.
• db-01 (storage primitives) — epoll/kqueue, zero-copy (splice/sendfile), and page-cache behavior from gw-01 build on the syscall-level intuition from db-01.