The Evolving WAN Architecture

I recently surveyed more than 100 IT professionals on the topic of wide area networking; my findings are in "The 2015 State of the WAN Report." I gave survey respondents a broad set of factors and asked them to indicate which ones play the biggest role in driving change in their company’s WAN. The factors cited most frequently include supporting real-time applications, increasing security, providing access to public cloud services, and reducing cost.

New technologies are emerging that enable a new branch office WAN architecture, one that promises to do a better job responding to these requirements than traditional branch office WAN architecture. Before I examine the new software-defined WAN, let's look back at some WAN history.

The Internet got its start in 1969 with the deployment of the ARPANET. The 20-year period that began around 1984 saw the deployment of four distinct generations of enterprise WAN technologies:

Unfortunately, there hasn’t been a major new WAN technology introduced into the market in over a decade. That fact, combined with the fact that organizations are continually retiring older WAN technologies such as frame relay and ATM, means that enterprise network organizations increasingly have to satisfy evolving demands on the WAN by using just two services: MPLS and the Internet.

The traditional approach to designing a branch office WAN is for each branch office to have either a T1 link or a set of bonded T1 links that provide access to a service provider’s MPLS network and to have one or more higher speed links at each data center. In this design, it is common to have all or some of a company’s Internet traffic be backhauled to a data center before being handed off to the Internet. One of the limitations of this design is that since the Internet traffic transits the MPLS link, this adds both cost and delay.

A number of vendors have released products that aim to overcome the limitations associated with  traditional branch office WAN architecture. I recently wrote a report entitled "The 2015 Guide to WAN Architecture and Design;" throughout this column, I'll refer to that report as The Architecture Guide. One of the guide's goals was to identify the key components of the emerging branch office WAN architecture and another goal was to create a side-by-side comparison of the emerging WAN solutions from a number of key vendors.

Figure 1: (Image: geralt/Pixabay)

To achieve those goals, The Architecture Guide included the description of a hypothetical company which operates a traditional WAN. The sponsors of the guide, Cisco, Silver Peak, Nuage Networks, Talari Networks, and Viptela, were asked to suggest how the hypothetical company should evolve its WAN..

As noted, in the traditional design, branch office connectivity is provided via T1-based access to a MPLS service. It has been possible for a while to use multi-pathing to distribute WAN traffic out of a branch office over a T1 link and an Internet link. However, relatively few network organizations implement multi-pathing in their branch office networks, in large part due to the difficulty of configuring the equipment.

All of the solutions that the sponsors presented fit the broad definition of a software-defined WAN (SD-WAN). As is the case with any software-defined network, a SD-WAN centralizes the network control function into a controller. The controller abstracts the user’s private network services from the underlying IP network and enables the operation of the user’s private network services via centralized policy. The controller also enables the automation of management tasks, such as configuration and provisioning.

Because of the ability of the controller to centralize policy and to automate configuration and provisioning, a common theme of the SD-WAN technologies was the promise to eliminate the traditional challenges associated with multi-pathing. One way that network organizations can leverage this functionality is to load balance traffic at each branch office over both MPLS and Internet links with the goal of reducing the capacity of the MPLS links and replacing the reduced MPLS bandwidth with relatively inexpensive Internet bandwidth.

An alternative approach  is to eliminate the use of MPLS and load balance traffic over multiple Internet links. In either case, the SD-WAN based approach to multi-pathing enables network organizations to respond to one of the  top factors driving WAN changes, as cited in my survey -- the need to cut costs.

There are, of course, differences in terms of how vendors leverage SD-WAN concepts in order to implement multi-pathing. In some products, the traffic is assigned to a link in a static manner -- the controller specifies which type of traffic uses which WAN link and unless there is a network failure or a change in policy, that allocation of traffic stays in place independent of the characteristics of the WAN links. Other products make the selection of the best end-to-end WAN path based on real-time traffic analytics, including the instantaneous end-to-end performance of each available network, the instantaneous load for each end-to-end path, and the characteristics of each application.

A more intelligent approach to multi-pathing is only one of the advantages of the emerging set of WAN solutions. In my next column, I will describe some of the others.