Introduction
EVPN is often touted as the next generation service layer for Service Provider networks. We'd contend that it shouldn't be considered as such, but rather, EVPN is the current technology and all other Layer 2 service technologies are legacy.
EVPN is a stable, standardised technology - EANTC
Legacy Protocols
So, what do I mean by legacy protocols. If you're deploying a Layer 2 protocol that doesn't signal MAC layer reachability via BGP, it's a legacy protocol. If your protocol uses standard layer 2 Broadcast Unknown Multicast (socalled BUM) traffic forwarding with MAC learning, it's legacy. Examples of this are Pseudowires (regardless of LDP or BGP signalled variant), VPLS, VPWS, etc. If it's a MAC learning protocol (or even without but attached to a software bridge - aka bridge domain) and doesn't signal MAC reachability via BGP, it's legacy.
So what?
Well, without BGP signalling, we have a few issues. Existing VPLS and related protocols cannot provide an all-active per flow redundancy. EVPN offers 4 different topologies, including all-active redundancy. This allows a CE to peer with two PEs and LAG the link where traffic is hashed per flow.
There are a number of requirements behind this technology, such as
- blocking BUM traffic to remote dual homed CE (Designated Forwarder Election),
- preventing flooded traffic from echoing back to multihomed device (Split Horizon)
- how to load-balance traffic towards a multi-homed device across multiple PEs when MAC addresses are learnt by only a single PE (Aliasing)
EVPN can handle all of these, whereas legacy protocols cannot.
Additional topologies
The Metro Ethernet Forum defines E-Line (Point-to-Point) and E-LAN (Multipoint-to-Multipoint) services, and this can, to an approximation, be delivered by the legacy protocols. But what about E-TREE (Rooted-Multipoint)? As RFC 8317 notes, the VPLS based ETREE mechanism of filtering at egress is a bad idea. To get around this, EVPN deploys an ETREE leaf Split Horizon Group label that ensures the leaves cannot communicate with each other.
Advanced mechanisms
If the above wasn't enough, EVPN has a few additional tricks up its sleeves. It can solve the following problems that legacy protocols cannot:
- How to inform other Leafs of a failure affecting many MAC addresses quickly while the control-plane re-converges (Mass MAC Withdraw)
- How to detect the correct location of MAC after the movement of host from one Ethernet Segment to another (MAC Mobility)
- How to route / how to go from one EVLAN Instance to another (IRBs, including symmetric and asymmetric)
Legacy integration and migration
As our managers and customer love to point out, our networks are brownfield. We only have to look at IPv6 to see the pain of a "forklift upgrade". Fortunately, EVPN has solutions. It can integrate with an existing VPLS network (EVPN Virtual Ethernet Segment) as well as support simple migration (Seamless VPLS to EVPN Migration). EVPN can also act as a Pseudowire Headend that aggregates Pseudowires.
EVPN ≠ next generation
So EVPN is not next generation but rather, the current generation. It's been widely deployed in Data Centres on top of VXLAN and is gaining widespread acceptance in Service Provider networks. It supports migration and interoperability with legacy protocols and delivers multiple useful features and topologies that are difficult or not possible with those legacy protocols. If you're interested, we have a suite of webinars where we go into great depth discussing these and other topics and we'd love to see you there.