3 Key WAN Architecture Considerations

Today, network organizations face a large and growing set of WAN architecture options. In my last column, I discussed one of those alternatives: where to locate key functionality. In this column, I'll discuss other WAN architectural alternatives and challenges facing network organizations.

Dynamic multi-pathing

Being able to load-balance traffic over multiple WAN links isn’t a new capability. However, in a traditional WAN, this capability was difficult to configure and the assignment of traffic to a given WAN link was usually done in a static fashion. One of the downsides of load balancing traffic in a static fashion is that the assignment of traffic to a given WAN link can’t change even if faced with adverse congestions such as a congested link.

There's now functionality available that enables dynamic load balancing over WAN links  based on a combination of policy and WAN link characteristics. One approach to leveraging this functionality is to dynamically load balance traffic over both MPLS and Internet links with the goal of reducing the capacity, and hence the cost, of the MPLS links and replacing the reduced MPLS bandwidth with relatively inexpensive Internet bandwidth.

An alternative approach is to use this functionality to potentially save even more money by load balancing traffic over multiple Internet links. To increase network availability, network engineers should strive to have diverse routing into their branch offices wherever possible. Having multiple Internet connections each  running over DSL is unlikely to have diversity of the first mile even if disparate ISPs are used. Having multiple Internet connections, one using DSL and the other using cable, will increase the likelihood of first-mile diversity.

To further increase the likelihood of first-mile diversity, network engineers should consider implementing Internet access over cellular either as an active link or as a backup.

Policy

There is a broad movement to implement a policy-based approach to all aspects of IT, including networking. Policies can be based on hierarchical system of rules designed to deal with the complexities of the environment and to manage the relationships among users, services, SLAs, and device-level performance metrics.

One way that policy can be implemented is at the application level. For example, if the performance of an application begins to degrade because the CPU utilization of a physical server hosting a virtualized network function (VNF)  used by that application becomes excessive, the VNF may be moved to a server with lower utilization, if in line with the application's policy.

As I alluded to in the discussion of dynamic multi-pathing, another way to implement policy-based networking is to control which WAN link application traffic transits based in part on centralized policies that indicate the business criticality of that application, among other characteristics.

Visibility

There are many tools that claim to provide network organizations with full visibility into a traditional WAN for troubleshooting problems related to network and/or application performance. However, whether it's deficiencies of those tools or the troubleshooting processes used by network organizations, survey data in my 2015 State of the WAN Report showed that less than one out of five network organizations has the visibility needed to effectively troubleshoot problems.

The adoption of new WAN architectures will further complicate the task of troubleshooting. For example, once dynamic load balancing is implemented, network organizations won’t be able to effectively troubleshoot the source of degraded application performance without knowing which WAN link supported the application’s traffic.

Evaluating new WAN solutions creates a variety of opportunities for network organizations. One of the opportunities is that by implementing a new WAN design, network organizations might be able to lower cost and increase availability. These new WAN solutions also create a variety of challenges, including whether or not they work, scale and actually deliver on their promises. Network teams need to ensure that as they explore new WAN alternatives, they evaluate the visibility provided by each of those alternatives.

Implementing a new WAN solution that lowers cost and increases network availability is clearly an important goal, but it’s not worth doing if it reduces the network organization's ability to troubleshoot.