Like the previous generations of mobile network standards (2G, 3G, 4G …), 5G is driven by the promise of faster and more efficient radio access technology that promises ubiquitous broadband (50Mbps everywhere), lower latency, seamless communication for mobility at higher speeds, and extreme scale. But the 5G architecture is also introducing a re-architecture of the core, incorporating new technologies and approaches that have emerged in the telco networking industry – software-defined networking (SDN), network functions virtualization (NFV), and mobile edge computing (MEC) to name a few. In addition, the architecture is required to support sharing infrastructure, providing various levels of “as-a-service” offerings – leading to complex architectures.
So how should we think about 5G? Faster speeds and greater density in the access? A bunch of new technologies that need to be put together? A network that requires complex architecture integration? A network that will satisfy an unbelievable range of scale and performance? In a way it is all of these, but we need to step back a bit from the trees to see the forest. Let’s focus on what operators are really trying to do and why.
Mobile broadband has had a significant impact on human behavior and business opportunity, and has essentially allowed the monetization of the Internet. For operators, the dilemma has been to undertake a succession of infrastructure modernizations to increase the speed and availability of mobile broadband, with the lion’s share of the resultant subscriber revenue going to an increasingly demanding ecosystem of over the top (OTT) providers and vertical industry players, rather than to the operators.
The future of network services for the consumer is bound up in economies of scale because only massive economies can hope to sustain service and application prices in a range where the mass market can pay. The future of services for the enterprise is however about new capabilities that require dynamicity, flexibility, and allow enterprises more control over their networks. Carriers will realize the benefits of the new modernizations and new infrastructure only when they offer experiences for people and organizations that they can’t get today, by providing the agile, flexible infrastructure required to sustain these futures.
The NFV movement has launched a transformation journey towards cloudification of the telco networks. Mobile broadband and 5G are guiding us on this journey to the cloud – but they actually encourage us to go beyond and complete the transformation of operators from being just connectivity providers to becoming digital service providers.
Beyond the Cloudified Network to the Digital Service Provider
5G is a trigger for requiring MEC because of the massive amount of data pouring in at the edge, enabled by the new high speed broadband that radio technologies deploy. New services made possible by this data flood can only be realized if timely local decisions can be made for mobile users and local storage buffers are available. This requires shared computing resources at switching centers and base stations. A parallel trend in landline networks, triggered by similar scale requirements, is resulting in plans for placement of compute facilities at the edge of the network in central offices and even in building basements. In addition to the advent of edge computing, there is a surge in the availability of compute and storage facilities further back in the network, where the advent of NFV is bringing data centers to the network. All of this computing, coupled with the associated management and orchestration required to operate it, will result in a distributed, dynamically configurable, compute matrix in the operator infrastructure.
5G creates the need to handle dynamic changes in network connectivity due to mobility at higher speeds and the ability to meet demand spikes that are mobile themselves. This accelerates the drive towards advanced programmability for the underlying networks. Several strata of network connectivity are required in the new software defined networks – we can think of overlay connectivity, function chaining, and path topology as a shorthand for understanding these strata. These connectivity strata all require control, in order to dynamically connect the applications, the software functions, and the compute matrix respectively. The control plane needs to be separated from the corresponding data/bearer plane, so that the capacity growth of the data/bearer and control traffic can evolve independently, and the routing can be optimized independently. The control required is typically local but needs to be globally informed and based on service-level policies. Coordination of control across strata is desirable, but the dynamicity at each topology stratum operates at a different time scale. So we require awareness across adjacent strata but not tight coupling. The same is true when we consider control across different parts of the operator network. A distributed and federated control infrastructure is the answer. These strata of network connectivity and the associated federated distributed control, will provide a programmable multi-layer topology in the communication service provider (CSP) infrastructure.
Finally the need for awareness across control layers plus the need for more granular and timely visibility into the network, fosters the ubiquitous deployment of information gathering and analytics in the network. The analytics required is both reactive and predictive, to be used for service assurance as well as for anticipating user needs. Analytics also provides the means to implement the loose coupling between the control layers mentioned earlier. Analytics, together with network control, a management and orchestration framework and the judicious use of automation, converts information into decisions and actions. We then have the means for automatically and dynamically responding to changes in infrastructure conditions and user behavior – resulting in a pervasive decision fabric in the provider’s infrastructure.
This new infrastructure, composed of a distributed compute matrix, a programmable multi-layer topology, and a pervasive decision fabric, is what a service provider needs to be able to change from being a connectivity provider to a digital service provider.