EVPN Type-5 Route with VXLAN encapsulation for EVPN-VXLAN
EVPN is a flexible solution that uses Layer 2 overlays to interconnect multiple edges (virtual machines) within a data center. Traditionally, the data center is built as a flat Layer 2 network with issues such as flooding, limitations in redundancy and provisioning, and high volumes of MAC addresses learned, which cause churn at node failures. EVPN is designed to address these issues without disturbing flat MAC connectivity.
VXLAN is an overlay technology that encapsulates MAC frames into a UDP header at Layer 2. Communication is established between two virtual tunnel endpoints (VTEPs). VTEPs encapsulate the virtual machine traffic into a VXLAN header, as well as strip off the encapsulation. Virtual machines can only communicate with each other when they belong to the same VXLAN segment. A 24-bit virtual network identifier (VNID) uniquely identifies the VXLAN segment. This enables having the same MAC frames across multiple VXLAN segments without traffic crossover. Multicast in VXLAN is implemented as Layer 3 multicast, in which endpoints subscribe to groups.
When a Bridge Domain (BD) is not L2 extended across Data Centers (DC), the IP subnet belonging to the BD is confined within a single DC. If all BDs within each DC network satisfy this requirement, there is no longer a need to advertise MAC+IP route for each tenant between data centers as host routes for the tenants can be aggregated. Thus the L2 inter-DC connectivity issue can be simply transformed to an inter-DC L3 IP prefix reachability issue.
Starting with Junos OS Release 17.1, the EVPN type-5 IP prefix route advertises the IP prefixes between the DCs. Unlike the type-2 EVPN MAC advertisement route, the EVPN type-5 IP prefix route seperates the host MAC address from its IP address and provides a clean advertisement of an IP prefix for the bridge domain.
Junos OS Release 17.1 also supports:
Inter-DC connectivity with VXLAN encapsulation for EVPN/VXLAN by using the EVPN type 5 IP prefix route. Each BD within a DC is not L2 extended. If EVPN/VXLAN is enabled between DC GW (Data Center Gateway) router and ToR while providing inter-DC connectivity, the spine, which acts as a DC GW router, is capable of performing L3 routing and IRB functions.
Inter-pod connectivity with VXLAN encapsulation by using the EVPN type-5 IP prefix route. The solution provided does not address the L2 extension problem when a BD is stretched across different pods. The spines that provide the inter-pod connectivity is able to perform L3 routing and IRB functions.