VXLAN, OVSDB, and BGP EVPN – 3 pillars of data center overlay
2021-03-26 | 5 min read
Overlay networking is a backbone of modern data centers
Hyperscale data centers need to handle the increasing demand of big data, cloud computing, server virtualization, and real-time applications. The data center network needs to scale up and down dynamically based on the growth of the workloads. To be able to connect applications running in different servers and adopt changes when the applications move, the networking infrastructure also needs to offer provisioning agility. A virtualized network overlay is extremely common to solve these challenges; enabling network segmentation, flexible workload placement, and dynamic provisioning based on demand.
Overlay is a technique to create virtual networks on top of a Layer 2 or Layer 3 physical network. These overlay paths route the traffic coming in and out of the data center and also within the virtual machines inside the data centers. Tunnels are used to transport payloads across an underlay (physical) network that has no visibility into or awareness of the traffic in those tunnels. Data center operators prefer overlay because it reduces the need for frequent modifications to the physical network while providing more agile deployment of networking resources.
Overlay encapsulation protocol – VXLAN
VXLAN is one of the most popular and effective overlay encapsulation protocols, capable of relaying Layer 2 traffic over IP networks. This enables creation of virtualized Layer 2 subnets or segments spanning physical Layer 3 networks. Existing VLAN-based segmentation has its limits in terms of smaller address space and in virtual machine (VM) mobility. VXLAN solves this limitation by enabling 16 million logical networks. The VXLAN-based network design logically isolates tenants and applications in a data center environment. While VXLAN-based overlays are a powerful tool, this comes with risks of network failure if not validated properly. Watch this demonstration about validating VXLAN using Keysight’s IxNetwork.
Enabling centralized provisioning of overlay - OVSDB
Quick and centralized provisioning of the overlay network is critical to support fast service delivery and application agility. This can be done using a central controller for distributing VXLAN Network Identifier (VNI) of the VM and mapping of VXLAN Tunnel End Point (VTEP) across the network. The OVSDB with hardware VTEP schema defines the structure of VTEP database. A successful implementation needs to validate the OVSDB Controllers and ability of the network in handling broadcast, unknown-unicast and multicast (BUM) traffic. Watch this demo to learn more about how IxNetwork OVSDB validation solution helps.
A control plane for distributed end host information – BGP EVPN
The VXLAN overlay mechanism does not change the flood and learn behavior of Ethernet protocol limiting in scalability and efficiency. The solution is to introduce a control plane to minimize flooding and facilitate learning. The control plane can distribute the end host information to the virtual tunnel end points. BGP EVPN is adopted as an extremely popular solution in data centers to carry the L2 MAC and L3 IP information over the control plane. But to create a large-scale multi-tenant network using BGP EVPN, network engineers need a validation plan. The IxNetwork BGP EVPN validation solution provides the knobs to execute a successful validation strategy. Check out How to validate EVPN VXLAN multi-tenancy and How to validate EVPN MAC mobility.