One of the use cases that Cisco VTS provides for is the automated provisioning of VxLAN overlays for L2 segments that spans top-of-rack (ToR) switches, or Points of Deployment (PODs).

Various deployments scenarios such as those associated with HA and VM mobility may require Layer 2 adjacencies between applications or servers connected to different ToRs residing on the same or different POD. The ToRs are interconnected via an IP network in a spine-leaf topology.

The Cisco VTS solution helps automate the provisioning of the VxLAN overlay between the ToR switches, which acts as VxLAN tunnel endpoints (VTEP). Not only is Cisco VTS able to perform the configuration of the ToR switch, it is also able to allocate resources such as VNID, local VLAN pools as well as multicast addresses which are assigned for the VNI.

Through the use of the Virtual Machine Manager (VMM) plug-ins, once the Tenant, Network, and VM are created, configuration that is required will be pushed down to the ToRs using the NCS component which is embedded in VTC. The IOS XRv component within VTC acts as a BGP EVPN route reflector and distributes host IP and MAC address reachability information learnt from each VTEP to other VTEPs. At the same time, the IOS XRv component within VTC advertises the host IP and MAC address reachability information to the BGP route reflectors using Multi-Protocol BGP (MP-BGP) in the EVPN based VxLAN deployment.

In traditional tree-based data center network design, there is a clear demarcation of the Layer 2 and Layer 3 boundaries. In these deployments, routing occurs at an aggregation point (e.g. core switch or router uplink). This causes a large amount of state information to converge on a pair of devices and may cause scale issues associated with large number of Layer 2 segments as in the case of data centers.

On the contrary in spine-leaf network design, if the leaf nodes were to implement routing by default and perform overlay bridging using VxLAN where required, the state information is distributed amongst leaf nodes and allows for an optimal and scalable deployment in the data center. This addresses the issues of scale problems associated with large Layer 2 broadcast domains as the leaf nodes, which are the top-of-rack switches have effectively terminated the broadcast domain.

Cisco VTS helps in the deployment of distributed gateway functionality by enabling the leaf nodes to perform routing by default and allows for the automated provisioning of VxLAN overlays on demand, eliminating the overhead typically associated with managing and configuring VxLANs.

In the case of VxLAN routing, a common VNI will be provisioned amongst different VTEPs to carry routed traffic. Routed traffic between the VTEPs are encapsulated in the VNI and longest prefix-match routing will then take place.

The Host routes for all known remote hosts are automatically installed at every VTEP and traffic towards these hosts will be forwarded over the VNI. Traffic to local hosts is provided by directly connected prefix since the host routes will not be present.

Cisco VTS may be used for the automated provisioning of the VNI that is used between the VTEPs to carry routed traffic. The IOS-XRv component can act as RR for the IP host routes as well as IP prefixes distribution among VTEPs/L3 gateways.

In a data center network, Data Center Interconnect (DCI) or Provider Edge (PE) routers provides external connectivity. If the PE routers do not support eVPN based VxLAN, they will need to be connected to the border leaf node.

In Cisco VTS, we support a two-box peering model between the border leaf node with one or more DCI or PE devices.

In VTS 1.5, a two-box peering model is supported for the DC-gateway with per-VRF BGP sessions between the DC-gateway and one or more external DCI/PE routers. In a future release, an integrated DC-gateway/PE solution will be supported.

To provide external connectivity to one or more virtual networks, the L3 VNI VXLAN segment will be extended up to the DC-gateway. The DC-gateway routing table will be populated with the L3 host IP routes from those virtual networks.