Course 45 - IE Data Center Network Design | Episode 4: Cisco Data Center Fabrics

Data Center Network Architecture: Multi-Site VXLAN EVPN and Chassis VirtualizationEpisode OverviewModern data centers increasingly depend on virtualization at both the network and hardware layers.Overlay technologies such as VXLAN BGP EVPN allow organizations to build scalable multi-site fabrics that support workload mobility, Layer 2 extension, and multi-tenant Layer 3 routing. At the same time, chassis virtualization technologies such as Cisco Nexus Virtual Device Contexts (VDCs) allow a single physical switching platform to operate as multiple logically independent network devices.In this episode, we explore both approaches through two interconnected topics:- Multi-site VXLAN BGP EVPN architecture- Cisco Nexus chassis virtualization with VDCsTogether, these technologies demonstrate how modern data center infrastructure can separate tenants, services, control planes, and failure domains while maximizing the capabilities of the underlying physical infrastructure.Part I: Multi-Site VXLAN BGP EVPN Design1. Understanding VXLAN EncapsulationTraditional Layer 2 networks can become difficult to scale across large data centers and geographically separated sites.Virtual Extensible LAN (VXLAN) addresses this limitation by encapsulating Ethernet frames inside UDP/IP packets, allowing Layer 2 connectivity to traverse a Layer 3 transport network.Conceptually:Original Ethernet Frame → VXLAN Encapsulation → UDP/IP Transport → VXLAN DecapsulationThe additional encapsulation introduces overhead, which must be considered when designing the underlying transport network.For the environments discussed in this episode, the additional VXLAN overhead can require jumbo-frame support across relevant data center interconnect paths.This is particularly important when supporting workloads that depend on Layer 2 adjacency, including certain virtual-machine migration scenarios.2. Combining Layer 2 Mobility with Layer 3 VRFsVXLAN is not limited to extending Layer 2 segments.Combined with BGP EVPN, it can provide a scalable control plane for both Layer 2 bridging and Layer 3 routing.This enables a fabric to support:- Layer 2 network segments.- Layer 3 tenant VRFs.- Workload mobility.- Network segmentation.- Distributed routing.- Multi-tenant environments.The resulting architecture can provide Layer 2 connectivity where mobility requires it while maintaining Layer 3 isolation between different tenants or application environments.3. Connecting Independent Data Center FabricsMulti-site designs introduce another challenge: how can separate VXLAN fabrics communicate without exposing their entire internal topology to one another?This is where border gateways become important.A border gateway can be deployed on an appropriate leaf or spine platform and act as a controlled boundary between independent fabrics.The gateway can:- Terminate VXLAN tunnels.- Exchange routes between sites.- Provide external connectivity.- Hide internal VTEP addressing.- Present a controlled routing boundary between fabrics.This creates an architectural separation between the internal overlay of each data center and the inter-site transport network.4. VTEP Address MaskingVXLAN Tunnel Endpoint (VTEP) addresses are normally used internally to identify tunnel endpoints.When multiple fabrics are interconnected, exposing every internal VTEP address across the entire environment can unnecessarily increase routing complexity.Border gateways can therefore provide a form of VTEP masking, allowing the inter-site control plane to operate through defined gateway endpoints rather than requiring every remote fabric to understand every internal VTEP.The conceptual architecture becomes:Fabric A → Border Gateway → Inter-Site BGP → Border Gateway → Fabric BThis creates a cleaner boundary between the independent VXLAN domains.5. Moving Beyond Legacy Multi-Site ExtensionsEarlier Cisco data center architectures used technologies such as OTV to extend Layer 2 connectivity between sites.Modern EVPN-based architectures can provide native multi-site capabilities within the VXLAN control-plane model.Depending on scale and physical geography, fabrics can be interconnected through designs such as:- Back-to-back border-gateway connections.- Dedicated inter-site links.- Long-distance superspine architectures.- Routed data center interconnects.The objective is to create a scalable interconnection model without turning the entire environment into one flat Layer 2 network.6. BGP and Multi-Site Control-Plane DesignBGP provides the routing foundation for many VXLAN EVPN deployments.Multi-site environments introduce additional considerations around:- Autonomous System Numbers.- R