Table of Contents
Introduction
Ethernet VPN, usually written as EVPN, is a modern technology used in large and complex networks to carry Layer 2 and Layer 3 connectivity over a Layer 3 core. It is especially important in data centers, large campuses, and service provider networks where you want flexible, scalable, and resilient connectivity between many sites or many racks, without creating a giant flat Layer 2 network.
In this chapter you will see what makes EVPN different from older VPN methods, how it uses a control plane instead of pure flooding, and why it pairs so well with VXLAN and other overlay technologies introduced in related chapters. You will not learn all of the internal MPLS or VXLAN details here, only what is specific to EVPN itself.
EVPN as a Control Plane for Ethernet
Traditional Ethernet relies heavily on flooding and learning. Switches flood unknown unicast, broadcast, and multicast traffic, and they learn MAC addresses by observing the source MAC on incoming frames. This works in small networks, but at the scale of large data centers or carrier networks it becomes inefficient and difficult to control.
EVPN changes this behavior by introducing a control plane for Ethernet learning. Instead of learning MAC addresses only from data traffic, EVPN uses a routing protocol, typically BGP, to distribute information about MAC addresses, IP addresses, and which devices are attached where. The MAC learning becomes driven by control plane messages, not just by the data plane.
In practice, each network device that participates in EVPN, for example data center leaf switches or provider edge routers, advertises which MAC and IP addresses it has locally. Other devices receive this information and build tables before any data traffic needs to be flooded. This reduces unknown unicast flooding, speeds convergence, and makes it easier to apply policies.
EVPN replaces pure data plane MAC learning with control plane learning carried primarily by BGP EVPN routes.
EVPN and Overlays
EVPN on its own is a control plane concept. It is often combined with an encapsulation technology that creates an overlay over an IP or MPLS core. The two most common combinations in practice are EVPN with MPLS and EVPN with VXLAN.
In an EVPN MPLS design, the EVPN routes learned through BGP tell the provider edge devices which pseudowires or labels to use to reach remote MAC or IP endpoints. The underlying network forwards labeled packets across the core using MPLS.
In an EVPN VXLAN design, which is very popular inside data centers, the EVPN routes tell leaf switches which VXLAN tunnel endpoints to use for specific MAC and IP destinations inside a given virtual network. Here the underlay is a regular IP network. The EVPN VXLAN combination allows the network to build multi tenant L2 and L3 overlays over a scalable Layer 3 spine and leaf fabric.
Although the encapsulation methods are different, the core EVPN idea remains the same. The control plane distributes reachability for hosts and networks, and the data plane just forwards encapsulated traffic based on that information.
EVPN Route Types and Information
EVPN extends BGP with special route types. These route types carry the information that switches and routers use to build Ethernet VPN services. At a high level, they describe what is reachable and how it can be reached.
For an absolute beginner, it is enough to understand that different EVPN route types carry different pieces of information:
| EVPN Route Type | High Level Meaning |
|---|---|
| Type 2 | MAC and IP reachability for individual hosts |
| Type 3 | Inclusive multicast and broadcast distribution info |
| Type 5 | IP prefix routes for Layer 3 VPN style reachability |
Type 2 routes are sometimes called MAC/IP advertisement routes. They say, in effect, that a particular MAC address, and often a matching IP address, is reachable via a specific EVPN node. This lets remote nodes forward unicast traffic directly, without flooding first.
Type 3 routes are used for broadcast, unknown unicast, and multicast, often abbreviated as BUM, distribution trees. They tell participating nodes how to reach all the others in the same broadcast domain.
Type 5 routes are used when EVPN is also providing Layer 3 VPN style routing between networks. In this case EVPN is not only emulating Ethernet segments, but also advertising IP prefixes so that routers can perform inter subnet routing within the same EVPN framework.
You do not need to memorize all encoding details. What matters here is that EVPN uses multiple specific route types to describe both Ethernet segments and IP subnets, and this flexibility is one of the reasons it is so powerful in modern networks.
Virtualization and Multi Tenancy with EVPN
EVPN is particularly useful when you need to support many tenants or many separate logical networks over one shared physical infrastructure. Each tenant can have its own isolated virtual network that might include Layer 2 segments and Layer 3 routing, while still sharing the same underlying switches and links.
To keep tenants separate, EVPN uses identifiers such as Ethernet VPN instances and virtual network identifiers. Inside one EVPN instance, hosts can be in the same broadcast domain, even if they are in different physical locations. Between EVPN instances, traffic is isolated, which allows secure multi tenancy.
In a data center, a common pattern is that each leaf switch attaches to servers from multiple tenants. The leaf tags each tenant’s traffic with the proper overlay identifier and uses EVPN routes to find the correct remote tunnel endpoints. The result is that a tenant sees its network as if it were one contiguous LAN, even though it is actually stitched together through many EVPN nodes and tunnels.
Active Active Multi Homing
Another key advantage of EVPN is better multi homing behavior. Multi homing means connecting a device, for example a server or a customer router, to more than one network device for redundancy. In older technologies, this often required special mechanisms or limitations, and many designs used only one active path at a time to avoid loops.
EVPN introduces an explicit concept of Ethernet segments and uses BGP signaling to coordinate multi homed links. A multi homed device attaches to a set of EVPN nodes that share an Ethernet segment identifier. These nodes coordinate which of them can forward traffic for that segment and how to handle failure of one of the paths.
With EVPN, it is possible to have both links from a multi homed device active at the same time. The EVPN nodes perform load balancing and fast convergence if one link or one device fails. This active active multi homing capability is one of the reasons service providers and large enterprises adopt EVPN for access and aggregation layers.
Reducing Flooding and ARP Traffic
In traditional large Layer 2 networks, broadcast traffic, especially ARP requests, can become a burden. Since switches do not know which port leads to a particular IP address, they must flood ARP requests to all ports in the VLAN. This can waste bandwidth and processing resources.
EVPN improves this situation with control plane learning and optimized ARP handling. Because EVPN nodes exchange MAC and IP mappings, they can often answer ARP queries locally without flooding them across the entire network. For example, a leaf switch that has learned a remote host’s IP and MAC from a Type 2 route can respond on behalf of that remote host.
Some EVPN implementations provide features like ARP suppression using the EVPN control plane database. The result is lower broadcast volume, more stable performance, and better scaling as the number of hosts grows.
EVPN significantly reduces broadcast and unknown unicast flooding by using control plane learned MAC and IP information, often enabling ARP suppression.
EVPN and Integrated Layer 2 and Layer 3 Services
One of the strengths of EVPN is that it can deliver both Layer 2 and Layer 3 services within a single framework. Earlier technologies often separated these functions into different mechanisms, which made designs more complex.
With EVPN, you can provide Layer 2 extension between sites, so hosts appear to be in the same LAN, and at the same time provide Layer 3 routing between subnets, all under EVPN control. The same BGP EVPN sessions advertise host MAC addresses, host IP addresses, and even IP prefixes that represent routed networks.
In a data center, this allows seamless integration between bridging and routing. For example, an EVPN VXLAN leaf can act as the default gateway for a subnet, perform routing between VXLAN segments, and still use EVPN to learn remote MAC and IP details. EVPN becomes the single control plane that ties together Layer 2 overlays and Layer 3 overlays.
EVPN in Data Center Fabrics
Although EVPN was first defined in the context of MPLS provider networks, its most visible success has been in data center fabrics, especially when combined with VXLAN. The typical design uses a spine and leaf IP underlay. All spines and leaves run a simple IP routing protocol in the underlay. On top of that, the leaves run BGP EVPN sessions and build VXLAN tunnels to each other.
In this role, EVPN provides several capabilities at once. It distributes reachability for end hosts, handles multi tenancy through separate virtual networks, supports distributed default gateways, and provides active active connectivity to servers. Compared to older approaches, such as large spanning tree based Layer 2 domains, EVPN fabrics scale further, converge faster, and offer more predictable behavior.
EVPN also fits well with automation. Since it is based on BGP and structured control plane information, configuration can be generated and maintained by templates and tools, which is important in large cloud environments.
Summary
EVPN is a modern control plane solution for Ethernet VPN services that brings routing protocol ideas into the world of MAC and IP distribution. By using BGP to advertise MAC and IP information, EVPN reduces flooding, enhances scalability, and enables active active multi homing. It supports both Layer 2 extension and Layer 3 VPN style routing, which makes it a central element in many data center and service provider designs.
In combination with encapsulation technologies like MPLS and VXLAN, EVPN creates flexible overlays for multi tenant and highly available networks. Its role in advanced enterprise and cloud networking is to provide a unified, efficient, and policy aware way to connect hosts and networks across a large, shared infrastructure.