Organizations often choose to offer services from different datacenters. The strategy described below guarantees the availability of those services and it ensures that the capacities of the datacenters are fully utilized.
In order to offer services from different datacenters, it is required to make an L2 link between them. This L2 is used to move virtual machines between the datacenters and to offer services in a high-availability configuration. Commonly used ways to connect datacenters are: L2 services of a provider, darkfiber, xWDM, VPLS, and technologies developed by manufacturers.
To guarantee availability, services are offered from multiple datacenters. It is important to limit the fault domain to the datacenter location. But, this is contrary to the use of L2 stretching where there is only one large fault domain. Problems within the first datacenter can cause disruptions in the second one. For example: ARP and Traffic flooding and spanning-tree problems. When there are more than two locations, it is certainly challenging to stretch L2 connections because the current techniques do not provide sufficient scalability.
Fortunately, these problems can be solved with the use of EVPN (Ethernet Virtual Private Network). With EVPN, it is possible to stretch L2 connections across multiple locations without adverse consequences.
EVPN offers a solution to shortcomings in traditional datacenter interconnect solutions. An L3 link is realized between datacenters. Within this L3 link, Multiprotocol Border Gateway Protocol Ethernet Virtual Private Network (MP-BGP EVPN) is used. This protocol is used as an underlay & control plane and it is responsible for the distribution of the learned MAC and IP-addresses between the various components. VXLAN is used as an overlay and data plane.
With EVPN, many current L2 shortcomings within the datacenter are resolved and modern, scalable, next-generation interconnects are built.