Networking Study Topics
1. Foundational Concepts
- VLANs (Virtual Local Area Networks): The basic method of logically segmenting a physical network into isolated broadcast domains using 802.1Q tagging.
- East-West vs North-South Traffic: The two fundamental traffic patterns in a data center — East-West flows between servers internally, North-South flows between servers and the outside world.
- MTU / Jumbo Frames: The Maximum Transmission Unit defines the largest packet a link can carry; Jumbo Frames (9000 bytes) reduce overhead for storage and overlay traffic.
- Bonding Modes: The different strategies (active-backup, balance-rr, 802.3ad/LACP, etc.) for combining multiple NICs into a single logical interface for redundancy or throughput.
- DNS / DHCP in Virtualized Environments: How name resolution and IP address assignment work when thousands of VMs and containers are created and destroyed dynamically.
- SDN (Software-Defined Networking): The paradigm of separating the control plane (decisions about where traffic goes) from the data plane (the actual forwarding of packets), managed centrally through software.
- BGP (Border Gateway Protocol): The routing protocol that glues the internet together, increasingly used inside data centers for spine-leaf fabrics and multi-cluster routing.
- IPv4 / IPv6 Dual-Stack: Running both protocol versions simultaneously on the same infrastructure to support legacy and modern workloads side by side.
2. Current State / VMware Baseline
- NSX-T (VMware NSX): VMware's SDN platform providing distributed switching, routing, firewalling, and micro-segmentation — the networking layer being migrated away from.
3. Physical Connectivity & Redundancy
- LACP (Link Aggregation Control Protocol): A way for a server and a switch to "agree" on grouping multiple physical cables into one fast, redundant link.
- LLDP (Link Layer Discovery Protocol): A "neighbor greeting" protocol where devices tell each other their name and which port they are plugged into.
- SMLT (Split Multi-Link Trunking): A method to split a redundant connection across two different physical switches so that if one switch fails, the other keeps the server online.
- ECMP (Equal-Cost Multi-Path): A routing strategy that lets data take multiple "equally good" paths at the same time to increase total speed.
4. Physical Design & Management
- Spine-Leaf Architecture: A physical layout named after the parts of a plant — Leaf switches (where servers plug in) connect to every Spine switch (the backbone) to ensure every server is exactly the same distance away from every other server.
- DCIM (Data Center Infrastructure Management): Software used to track the "physical world" of the data center — power, cooling, rack space, and cable paths.
5. Virtualization & Overlays (The "Software" Layer)
- CNI (Container Network Interface): The standard "plug" that allows Kubernetes (OpenShift) to connect containers to different types of networking software.
- Multus (Multi-Network CNI): A meta-CNI plugin that allows a single Pod or VM in Kubernetes to attach to multiple networks simultaneously — critical for OVE where VMs need access to both the cluster network and external VLANs.
- OVS / OVN (Open vSwitch / Open Virtual Network): OVS is the switch made of software; OVN is the brain that controls many virtual switches across many servers.
- VXLAN (Virtual Extensible Local Area Network): A way to create a "Virtual LAN" that can stretch across different physical buildings by hiding it inside standard IP packets.
- GENEVE (Generic Network Virtualization Encapsulation): A "next-generation" version of VXLAN that has a flexible header (like a whiteboard) where the system can write extra notes about security or routing.
6. Advanced Data Paths
- SR-IOV (Single Root I/O Virtualization): A hardware feature that lets a single physical network card "slice" itself into many virtual ones that plug directly into VMs for maximum speed.
- IOMMU (Input-Output Memory Management Unit): A security chip feature that makes sure a VM only touches its own assigned piece of hardware and can't "spy" on other parts of the server's memory.
- DPDK (Data Plane Development Kit): A set of software libraries that lets a program grab network data directly from the card, bypassing the slow parts of the operating system.
7. Routing & Security
- DVR (Distributed Virtual Routing): Instead of one central "router" in the data center, every single server has its own "virtual router" built-in to handle traffic locally.
- VRF (Virtual Routing and Forwarding): Like a "Virtual Router" inside a physical one, allowing you to have multiple separate routing tables that don't talk to each other.
- eBPF (Extended Berkeley Packet Filter): A tiny, ultra-fast "computer inside the computer" that lets you run security and networking programs directly in the core of the Linux system.
- Micro-segmentation: The practice of applying firewall rules at the individual VM or Pod level rather than at the network perimeter — every workload gets its own security boundary.
- Network Policies (Kubernetes): The Kubernetes-native mechanism for defining which Pods/VMs can talk to which other Pods/VMs — the OVE equivalent of NSX firewall rules.
- QoS (Quality of Service): Mechanisms to prioritize certain types of traffic (e.g. storage replication, live migration) over others to guarantee performance under load.
- VPN / IPsec Tunneling: Encrypted tunnels between sites or clusters for secure communication over untrusted networks.
8. Platform-Specific Networking
- Kubernetes Networking Model: The fundamental rules — every Pod gets a real IP, every Pod can reach every other Pod without NAT, and Services provide stable endpoints.
- OpenShift Routes / Ingress Controllers: How external traffic enters the cluster and reaches the correct workload — the North-South entry point in OVE.
- MetalLB / Load Balancing: Load balancer implementations for bare-metal Kubernetes clusters that assign real IPs to Services — replaces cloud provider LBs.
9. Network Observability
- Network Flow Monitoring: Tools and techniques for capturing, analyzing, and visualizing traffic flows across the data center to detect anomalies and plan capacity.