While the rather broad topic of software defined networking, or SDN, still means different things to different people, one specific component has risen above the fog of ambiguity and can now claim to be mainstream. I’m referring, of course, to SD-WAN.
One obvious indicator of SD-WAN going mainstream is that many familiar vendors, including Cisco, VMware and Riverbed, have either acquired or assisted in the development of their own branded solutions. To me though, the idea of being mainstream is the simple fact that SD-WAN solutions solve an unambiguous problem. That being, how to best utilize all available bandwidth on offer to a site, large or small, reliably, at the lowest cost, and very importantly…with little OpEx overhead to already overworked network operations staff.
With cloud based SD-WAN solutions becoming the norm, OpEx savings are immediately realized. Of the cloud based solutions, Versa Networks is emerging as a very attractive choice.
Versa Networks is designed around the core fundamental of SDN, that is, the complete separation of the control-plane (Versa Controller) from the data-plane (Versa FlexVNF software installed on bare metal or as a VM at the branch). The Versa Controller is one of three components that make up the head end of the overall design. The other two being the Versa Analytics Engine, and the management UI, the Versa Director.
Since we’re talking about SDN here, and they’ve branded themselves Versa Networks, we as network engineers should be interested in how they go about piecing the software defined topology together, right? I thought you might agree! If you’ve got an understanding of the basics of service provider functionality, including virtual routers and multi-protocol BGP, then you’re good to go. A Versa Networks branch device, deployed in a multi-tenancy environment (yes, multi-tenancy is supported too!), spins up the following three types of VRs.
The Transport VR
The transport VR on a branch appliance running FlexVNF software owns the actual WAN link, or the multiple WAN links, assigned to it. Again, based upon SDN, we’re talking about the concept of over and underlay networks. The transport VR is considered the underlay network. It’s there to support the dynamics of the overlay networks that ride over the top of it. It can run a routing protocol if necessary, while providing Internet access.
The Control VR
The control VR is spun up and tied to a tenant. Each tenant configured on a Versa Networks branch gets their own control VR. Being the control plane element, as you’d expect, this VR is considered the brains of the device. It runs MP-BGP with the Versa controller, sending and receiving IPv4 and IPv6 prefixes unique to that tenant. By design, the peering of branch to controller is internal, or I-BGP. For this reason, to satisfy the rules surrounding I-BGP lacking a full mesh of devices, the controller acts as a BGP route reflector. Control VR to control VR communications is provided by IPSec or VXLAN tunnels.
The LAN VR
A LAN VR is also spun up and bound to each unique tenant. In this case, the branch devices LAN interface, or interfaces, are tied to this VR. Going back to the earlier reference to service provider topologies, the LAN VR is of the type virtual routing and forwarding. The LAN VR is paired to the control VR, which as described above, is responsible for sending the tenant routes to other devices. Acting like a service provider PE device, MP-BGP using route-targets and route-distinguishers, identifies tenant specific prefixes as they’re announced via the control VR.
Fundamentally, for traditional network engineers, that’s it. Keep in mind, while this might appear complex, with configuration done via the Versa Networks Director UI, most of this is done “under the hood”, through easy to use templates.
Of course, the power of the Versa Director UI really comes in to play when it’s time to set up policies that make SD-WAN truly “software defined”. Traffic engineering via application ID and QoS, to name just a couple of options, offer up numerous possibilities for cost savings and traffic optimization to the network operations team. So many in fact, we’ll save that topic for another time.