Skip to content

Networking

On-premises networking is fundamentally different from cloud networking. There is no VPC wizard, no managed load balancer, and no automatic DNS to rely on. You must design the physical topology, configure the switches, and run the protocols that make Kubernetes networking function efficiently and securely on bare metal.

In a cloud environment, software-defined networking abstracts away the complexities of the underlying physical infrastructure. On bare metal, you are responsible for bridging the gap between the physical data center network and the overlay networks used by your clusters. This requires a deep understanding of Layer 2 and Layer 3 constructs, routing protocols like BGP, and hardware-level considerations such as MTU and VLAN isolation.

This section covers the end-to-end networking stack required to run Kubernetes in your own data center. You will learn how to design a resilient spine-leaf architecture, implement dynamic routing, provision highly available load balancers without cloud provider APIs, and secure cross-cluster communication using service meshes and automated certificate authorities.

ModuleDescriptionTime
Module 3.1: Datacenter Network ArchitectureSpine-leaf topology, ToR, L2/L3, MTU, VLAN design60 min
Module 3.2: BGP & Routing for KubernetesBGP peering, Calico BGP, route reflectors, multi-site60 min
Module 3.3: Load Balancing Without CloudMetalLB, kube-vip, HAProxy/Keepalived60 min
Module 3.4: DNS & Certificate InfrastructureInternal DNS, split-horizon, cert-manager with Vault CA45 min
Cross-Cluster NetworkingInter-cluster routing, network policies, and stretched overlays45 min
Service Mesh on Bare MetalIstio, Linkerd, mTLS, and ingress gateway architectures60 min