NL001-Introduction to VxLAN

There has been a change in the trend. Historically, networks were built for applications, but today, applications determine how networks should be designed. There are several demands associated with an application, such as clustering of servers, mobility of workloads, cloud integration, and low delays with efficient forwarding. 

Therefore, network engineers like us build two-tier fabrics that are called CLOS fabrics, which are made up of leaf and spine switches. We translate application requirements into network architectures using technologies for multi-tenancy, Layer 2 extensions, and mechanisms for scaling networks, like VxLAN EVPN. VxLAN stands for Virtual Extensible LAN and EVPN means Ethernet Virtual Private Network. 

VxLAN is a standards-based protocol, not a proprietary one. There are multiple vendors who use this technology, and this blog is focused on Cisco-based implementations. With VxLAN, layer 2 traffic can be carried over layer 3 infrastructure. By encapsulating the Layer 2 frame with an extra VXLAN header, VXLAN uses a MAC-in-UDP (MAC-in-UDP) encapsulation technique. 

The VXLAN header is an 8-byte structure consisting of a 24-bit unique network ID (VXLAN Network Identifier (VNI) and several reserved bits. Using the VNI, Layer 2 segments are identified and isolated from one another at Layer 2. When all 24 bits of VNI are used, VXLAN is capable of supporting 16 million segments of local area networks. 

There are a range of valid VNI values between 4096 and 16777215. Moreover, IP/UDP encapsulation allows each LAN segment to be extended across Layer 3 networks by utilizing Layer 3 Equal-Cost Multipath (ECMP).