MP-BGP EVPN VXLAN is a widely adopted overlay networking technology that delivers scalable, efficient, and flexible network virtualization. By combining Multiprotocol BGP (MP-BGP), Ethernet VPN (EVPN), and VXLAN, it solves the key challenges of traditional Layer 2 networks in modern data centers.
Core Components of MP-BGP EVPN VXLAN #
- MP-BGP: A routing protocol used to exchange routing information across Autonomous Systems (AS).
- EVPN: A Layer 2 VPN technology that extends Ethernet segments over a Layer 3 underlay network.
- VXLAN: Encapsulates Layer 2 traffic (overlay network) within a Layer 3 IP network (underlay network), enabling large-scale virtualized networks.
Understanding Underlay and Overlay Networks #
A stable underlay network is the foundation for MP-BGP EVPN VXLAN. Typically, a Spine-Leaf or Clos topology is used with a routing protocol like OSPF or BGP. This design provides:
- Equal Cost Multipathing (ECMP) for load balancing
- High bisectional bandwidth
- Fault tolerance and redundancy
Leaf switches can also use MLAG or vPC for host connectivity, enhancing network reliability.
VXLAN uses Virtual Tunnel Endpoints (VTEPs) to encapsulate Layer 2 frames for transport across the underlay. Each VXLAN segment is identified by a VXLAN Network Identifier (VNI). For example, VLAN 100 frames can be encapsulated with VNI 10100, routed over the IP network, and decapsulated at the destination VTEP.
VXLAN Encapsulation Steps #
- Source VTEP removes the original VLAN tag.
- VXLAN header (VNI) is added to the frame.
- Destination VTEP decapsulates, restores the VLAN tag, and delivers the frame.
Note: All VTEPs must maintain consistent VNI-to-VLAN mappings across the network.
Challenges of Early VXLAN Deployments #
Initially, VXLAN adoption was limited due to inefficient MAC address learning. Unlike traditional Layer 2 switches that use flooding for unknown unicast and multicast (BUM) traffic, VXLAN needed multicast or head-end replication (HER) for BUM traffic. These methods did not scale well for large data centers, requiring frequent remote MAC re-learning.
Integrated Routing and Bridging (IRB) and Distributed Gateways #
VTEPs can perform Layer 3 routing with IRB, unifying gateway IP and MAC addresses across the network.
- Asymmetrical IRB: Forward and return paths differ. Can lead to inefficient MAC learning and scaling issues.
- Symmetrical IRB: Uses a dedicated Layer 3 VNI per VRF. Forward and return paths are identical, enabling better scalability.
Distributed Anycast Gateway: Configures the same gateway IP on all Leaf switches. Each Leaf serves as the default gateway for its directly connected hosts, reducing latency and improving efficiency.
MP-BGP EVPN: Solving VXLAN Limitations #
EVPN introduces a control plane for VXLAN using MP-BGP (RFC 7432, RFC 8365), replacing flooding-based MAC learning with proactive MAC/IP advertisement:
- Leaf switches advertise connected MAC and IP addresses via BGP.
- Spine switches synchronize this information across the network.
- Reduces BUM traffic and improves network scalability.
Key EVPN Concepts #
- Route Distinguisher (RD): Ensures uniqueness of routes across VPNs.
- Route Target (RT): Controls import/export of routes between VRFs.
Three-Layer Architecture #
- Underlay Network: Spine-Leaf topology with ECMP and high redundancy.
- Data Plane: VXLAN encapsulation for Layer 2/3 transport.
- Control Plane: MP-BGP EVPN distributes MAC/IP info using RD/RT for accurate endpoint mapping.
MP-BGP EVPN VXLAN Deployment Workflow #
Step 1: Establish the underlay network using OSPF or BGP on Spine-Leaf links. Create VTEPs on Leaf switches bound to loopback interfaces.
Step 2: Enable the MP-BGP EVPN address family, establish BGP neighbors via loopback IPs, and configure eBGP-multihop=3 for resilience. ECMP and P2P links optimize traffic distribution.
Step 3: Leaf switches advertise MAC/IP routes via BGP EVPN. VXLAN encapsulates Layer 2 frames for transport. Local MAC tables, BGP EVPN routes, and ARP caches handle most traffic, minimizing flooding.
Conclusion #
MP-BGP EVPN VXLAN provides a scalable and efficient overlay network solution for modern data centers. By combining:
- Stable underlay networks (Spine-Leaf topology)
- VXLAN encapsulation for Layer 2 traffic
- EVPN with MP-BGP for proactive MAC/IP distribution
it overcomes traditional VXLAN limitations, supports multi-tenancy, reduces flooding, and ensures high performance and interoperability across large-scale deployments.