VXLAN Overview

#

• Ethernet in IP/UDP encap port 4789
• VTEP
• NVE
• NVI

VXLAN EVPN Basics

#

• Leaf/Spine is ideal for east/west traffic flows
• VXLAN requires a mechanism to know which end hosts are behind which VTEP
  • This allows the VTEP to build the location-identity mapping database
  • Identity - identifies an end host using its IP, MAC, etc
  • Location - identifies the VTEP responsible for encap/decap tunnel traffic for that end host
• VXLAN Flood & Learn suffers from the scalability issue because it requires flooding the traffic to learn about end hosts behind a VTEP
• The use of MP-BGP and Ethernet VPN (EVPN) solves the scalability issue in VXLAN F&L by minimizing flooding
• The BGP L2VPN EVPN address-family allows the host MAC, IP, network, VRF, and VTEP info to be carried over MP-BGP
• The underlay network provides the reachability information to reach the VTEP, while the overlay control protocol (MP-BGP EVPN) distributes end-host info
• When a VTEP learns about a host on its local segment (e.g., ARP, DHCP), BGP EVPN distributes and provides this information to other BGP EVPN speaking VTEPs within the underlay network
• As long as the source VTEP continues to detect a host behind it, no EVPN update message is sent out, so other VTEPs don’t need to age out any remote host info
• For BUM traffic forwarding, the sourcing VTEP is required to send multi-destination traffic to mutiple VTEPS, either using IP multicast or ingress/head-end replcation
• MP-BGP EVPN has different route types
  • Cisco’s implementation of VXLAN EVPN uses types 2, 3, and 5

EVPN Route Types

#

• Type-2 Mac/IP Advertisement: provides end-host reachability info; MAC (mandatory) and host (/32) IP address (optional)
• Type-3 Inclusive Multicast Ethernet Tag: is used to create the distribution list for ingress replication
• Type-5 IP Route: provides IP prefix and length advertisement in EVPN

VXLAN BGP EVPN RD & RT

#

• VRF allows overlapping IP addresses with isolated routing domains
• MP-BGP uses the RD to differentiate between routes stored in the BGP tables
  • RD is 8-bytes (64-bits) and composed of 3 fields
  • Assign RD to each VRF and add RD to IP to maintain uniqueness among identical routes in different VRFs
• MP-BGP also uses a Route Target, which is an extended BGP community places on a route to control the import and export of BGP prefixes between VRFs
  • By using RT’s, VRF routes are exported from BGP VRF into VPN address-family and vice versa
  • RT is 8-bytes with prefix:suffix notation
• For simplification, Cisco provides automated derivation of RDs and RTs
  • For RD, the format is RID:VRF-ID (RD, 10.0.0.11:3)
  • For RT, the format is ASN:NVI (RT, 65501:50001)

VXLAN BGP EVPN Components

#

• IGP underlay routing
  • OSPF, EIGRP, IS-IS
• Multicast Underlay Routing
  • PIM sparse mode
  • PIM Anycast RP or Bidir Phantom RP on spines
• BGP Underlay control plane
  • Spines should be route reflectors
• Advertise MP-BGP EVPN EVPN Routes
  • MAC to L2 NVI to VTEP mapping
  • IP to L3 VNI to VTEP mapping
• VXLAN data plane encapsulation

VXLAN Packet Structure

#

• Inner header
  • Applies to original L2 frame
• Outer header
  • VXLAN header
    • VXLAN flags
    • Reserved bits
    • VNI
  • Outer UDP header
    • Source port
    • VXLAN port UDP 4789
    • UDP Length
    • Checksum
  • Outer IP header
    • IP Header Misc Data
    • Protocol 0x11 (UDP)
    • Header checksum
    • VTEP’s source IP
    • VTEP’s dest IP
  • Outer MAC header
    • Destination MAC (NH address)
    • Source MAC (VTEP address)
    • EtherType (0x0800)
• VXLAN encapsulation added 50 bytes of overhead
  • 20 bytes + 8 bytes + 8 bytes + 14 bytes = 50 bytes
  • Add another 4 bytes if dot1q is used

VXLAN EVPN Endpoint Learning and Forwarding

#

VXLAN EVPN L2 Forwarding

#

VXLAN EVPN L3 Forwarding

#

VXLAN EVPN Configuration

#

VXLAN EVPN Underlay Configuration

#

VXLAN EVPN Overlay L2 Bridging Configuration

#

VXLAN EVPN Overlay L3 Routing Configuration

#