Software Defined Networks (SDN) / SD-WAN

In simplistic terms SDN separates the software from the hardware. The hardware may be white box (off the shelf computer systems) or gray box (proprietary hardware that is opened to other vendors Operating Systems). SDN is separated into numerous layers like Device Drivers with access interfaces, layers where computational work is accomplished and one or more overlying layers that tie numerous systems together and feed up and down the chain the exchange of information.

Except for the edge devices (white or gray boxes) the programs, matrix, and computational array exist in Virtual Machines (VM). A VM is created for each service or program. Resources are allocated from a VASO or SESO stacks. The necessary information is then pushed out to the far end element.

One of the services provided the SDN stack is Software Defined Wide Area Network (SD-WAN). There is a great deal of “marketing” around SD-WAN . Things like “replace your slow 1.5mb MPLS circuit and SD-WAN is almost as good”.  The “MPLS circuit” is actually a product sold by VZ and ATT many years ago. It allowed multiple protocols to be transported over a T1 or DS3. Most importantly MPLS is a protocol and not tied to any speed or transport function. 

SD-WAN can be deployed over any type of circuit but most often is used with two Direct Internet Access links from two different carriers. A signaling protocol runs between the endpoints and communicates to the SD-WAN VM the best path for the upcoming packets thus, Software Defined Network. Each client has its own SD-WAN  VM.

Routing loops, collisions and broadcast storms are eliminated with EVPN services running on top that make the two far end physical circuits look like one end-point.

SD-WAN can make sense when cost or availability of MEF circuits makes them a bad choice. SD-WAN may not be applicable when the circuits, throughput, and reliability are mission critical. network engineering