By: Jonathan Homa

The telecommunications industry is quickly reaching an inflection point that is forcing communications service providers to rethink their approach to network architectures. First and foremost, 5G and fiber-to-the-premise buildouts are fueling massive increases in access and transport bandwidth. On the consumer side, there is rising demand for increasingly sophisticated services with defined performance characteristics such as high capacity and guaranteed ultra-low latency performance. These services include cloud gaming, telemedicine, and symmetrical home working, which in turn are fueling even more bandwidth demand. Finally, service providers are experiencing competition from new sources such as webscale content providers like AWS, Azure, and Google Cloud, which are slowly chipping away at provider revenues by offering Internet and cloud connectivity services to businesses.

With how networks are designed today, services are delivered in a pre-defined, non-programmable way. This approach takes an overly simplistic view of services, leaving providers unable to respond to increasingly complex service demands and a competitive environment.

These technological and market changes, however, are causing service providers to consider how they can optimize today’s networks to meet tomorrow’s demand—while also capturing new customers and reducing churn to new competitors in the space who are able to offer these services more quickly. For providers that are serious about their ability to navigate these factors to remain competitive and meet customer demand, it’s time to consider re-engineering the network to be services-driven, and step away from the transport-optimized approach of today.

The Transition to Services-Defined Networking

At the core of a services-driven approach is the focus on the services that providers deliver, rather than on the network plumbing required to deliver the services. Services dictate which network resources are used, how they connect, and the performance required from these connections. The network then instantiates resources and dynamically adapts connectivity to meet these needs.

The challenge of course is understanding how to design a network that is capable of supporting different performance needs. A single, uniform network that handles all service types would require massive over-engineering and be unreasonably expensive. The alternative—to build and maintain separate networks for each service type—is both incredibly intensive and unreasonably expensive, and would hinder providers’ abilities to introduce new service types in the long term.

The only feasible way to design this network is to virtually slice it through a method called network slicing. In this method, a single network offers shared network resources that can be optimized on a service-by-service basis. To create the dynamic control required by this kind of network, a combination of soft and hard network slicing must be implemented.

Hard slices allocate resources through their own dedicated traffic paths, keeping the services they carry fully isolated from others. This ensures that mission and business critical applications are shielded from competing demands on the network resources they use and are guaranteed a certain level of performance. Soft slices, on the other hand, usually compete for resources, such as their position in a buffer queue. As a result, services running over soft slices can only be engineered to have an average level of performance, typically called “best effort,” with performance rising and falling whenever the demand-to-bandwidth ratio for these resources is adjusted.

On a technological level, these networks will require a fully integrated programmable transport layer, composed of both packet and optical capabilities. Service-oriented operations software will provide the intent-driven linkage between the service requests coming from the orchestration layer and the configuration required in the transport layer. Together, technologies and techniques like virtualization, orchestration, automation, analytics, and cloud will create a programmable network that enables network as a service. This will power the transition of a static network to an agile one that proactively ensures the services delivered across the network meet service