Designing & Securing Multi-Tenant Runtime Environment at the New York Times
summary
The New York Times builds a centralized multi-tenant EKS platform across two regions, automating cluster provisioning with Terraform and Helm, removing the AWS VPC CNI and kube-proxy and installing Cilium (eBPF) CNI for kernel-space networking; observability uses Hubble, traffic control uses Istio with east–west gateways, and policy enforcement uses OPA Gatekeeper constraints (including custom constraints for IAM-for-ServiceAccount mappings). Clusters scale to hundreds and are designed to reach thousands of nodes; the setup enforces namespace isolation, resource limits, and service-to-service mTLS successfully.
AI-written summary · about
chapters (17)
- 0:00Start
- 2:15Developer Journey
- 2:59Innovation Cost
- 4:14Internal Developer Platform
- 5:39Cloud Architecture
- 7:44Design Considerations
- 9:04Multi-Tenant Architecture
- 9:48Runtime Components
- 11:28Performance
- 13:10Observability
- 14:37Install Script
- 16:28Tenant Onboarding & Isolation
- 18:23Tenant RBAC Example
- 24:17What is a Service Mesh
- 26:02Service Mesh Setup
- 26:37Multi-Tenant Setup
- 27:25Takeaways.
related talks
28mDesigning and Securing a Multi-Tenant Runtime Environment at the New York Times
28mWhat We Learned Designing & Securing a Multi-Tenant Developer Platform at The New York Times
24mScaling Argo Security and Multi-Tenancy in AWS EKS at the New York Times
26mMulti-Tenant vs Micro-Clusters in Kubernetes-Driven Platform Engineering
26mTwo sides of the platform coin with operator's and customer's voices
05mLightning Talk: What Makes a Good Multi-tenant Kubernetes Solution?
39mThe Multi-Tenant Supercluster Approach
29mManaging a Multi-Tenanted Kubernetes Cluster in Production