Network Overlay Technologies for Modern Virtualized Data Center Architectures

Modern virtualized data‑center architecture must satisfy requirements that accelerate application deployment and support DevOps, including scaling forwarding tables and subnets, layer‑2 expansion, virtual device mobility, forwarding‑path optimization, and multi‑tenant virtual networking on shared physical infrastructure.

Although network overlay is not a new concept, interest has risen because it can meet many of these requirements. New encapsulation frame formats built for data centers—such as Virtual Extensible LAN (VXLAN), Network Virtualization using Generic Routing Encapsulation (NVGRE), Transparent Interconnection of Lots of Links (TRILL), and Locator/ID Separation Protocol (LISP)—have driven this interest. An overlay network is a virtual network formed by interconnected nodes that share the underlying physical network, allowing deployment of applications with specific topologies without altering the physical network (see Figure 5).

Figure 5 Network Overlay Concept

Benefits of network virtualization overlay include:

Optimized device functionality: Overlay networks allow separation and specialization of device functions based on their location. Edge or leaf devices can optimize functions and protocols using endpoint state and scale, while spine or core devices can optimize based on link‑state updates, achieving fast convergence.

Structural scalability and flexibility: By focusing on scaling edge devices, overlay technology enables network expansion without requiring spine and core devices to store endpoint host information in their forwarding tables.

Overlapping addressing: Overlap techniques let virtual network IDs have unique scopes, allowing MAC and IP address overlap between tenants. Overlay encapsulation also separates the address space of the underlying infrastructure from tenant address spaces.

This document reviews several spine‑and‑leaf designs Cisco has offered recently, current designs, and upcoming designs intended to meet modern virtualized data‑center requirements:

Cisco® FabricPath spine‑and‑leaf network

Cisco VXLAN flood‑and‑learn spine‑and‑leaf network

Cisco VXLAN MP‑BGP EVPN spine‑and‑leaf network

Cisco Massive‑Scale Data Center (MSDC) layer‑3 spine‑and‑leaf network

Each section outlines the most important technical components at the time of writing—encapsulation, endpoint detection and distribution, broadcast/unknown‑unicast/multicast forwarding, underlay and overlay control planes, multi‑tenant support—and discusses common designs and design considerations such as layer‑3 gateways.

The article concludes with a reference link (http://jiagoushi.pro/node/1033) and invitations to join related knowledge communities and social platforms.