Building a high-performing programmable network
AT&T has embarked on the journey toward realizing an open, programmable network. This transformation marks a pivot from one-size-fits-all, best-effort mobile broadband, to a new era of performance-based business models built upon differentiated connectivity and open network infrastructure.
Key insights
High-performing programmable mobile networks exhibit:
- Openness: Open ecosystem, development tools, interfaces and technology.
- Agility: Incorporating new and innovative technologies and features rapidly.
- Optimization: Continuous use of predictive AI/ML-based analytics, improving maintenance, performance and user experience.
- Reliability: Proactive resolution of network issues before they occur, to provide low latency and a highly reliable service for critical applications.
- Quality: Consistent user experiences while ensuring security and privacy.
AT&T is improving its network to meet the growing demands of users, manage increased complexity without compromising TCO, and provide a competitive edge in the market.
Today, the AT&T nationwide 5G network covers more than 300 million people in nearly 24,900 cities and towns across the US. AT&T 5G is driving the next wave of innovation for consumers, businesses, first responders and government agencies with fast, reliable and secure connectivity.
The transformation to an open, programmable network is driven by the increasing demand for tailored network services that can meet the diverse and dynamic needs of users, applications, devices and enterprises. Examples include: high upload speeds at major events for video streamers, low latency for cloud gaming and guaranteed characteristics for business-critical processes. As part of the network evolution, AT&T will utilize intelligent automation to enable its network to be more flexible, efficient and responsive.
This article was written in cooperation with multinational telecommunications service provider AT&T, which has the largest network in North America.[1]
Enabling differentiated connectivity through programmable networks
Programmability in radio access networks (RAN) is a pivotal technology advancement, driving the shift toward more adaptive, efficient and differentiated networks. By integrating intelligent, programmable capabilities directly within the RAN layer through smart applications (rApps) and a service management and orchestration (SMO) framework, the delivery of optimized network performance and differentiated connectivity is made possible. Differentiated connectivity is the ability to offer a tailored set of performance characteristics like throughput, latency and jitter on both downlink and uplink through network slices. The characteristics are matched to an application’s requirement in a given service area to provide a predictable experience. Monitoring via observability capabilities is required to ensure those characteristics are delivered. For example, not all types of video traffic have the same requirements:
- “best-effort” is good enough for casual viewing
- for 4K video viewing, a volume-centric offering that can support large amounts of data would be more appropriate
- live camera feeds require latency-centric offerings
- for live broadcasting, a fully dedicated service would be required, supporting both high data volumes and low latency in the uplink
The flexibility within the network to meet distinct requirements will spur new innovations and meet the various dynamic needs of consumers and enterprises. Different types of connectivity characteristics could be requested via a quality on demand (QoD) API or via user equipment route selection policy (URSP) from the device.
RAN data enrichment for network optimization
The evolution of 5G has brought diverse use cases with varied performance demands. Programmability facilitates the management of such complex networks by enabling the tuning of the RAN behavior through standard interfaces defined in the O-RAN Alliance. These interfaces help developers to build portable applications (rApps using the R1 interface), manage behavior in the RAN elements (using the A1 policy interface), as well as perform operations, administration and maintenance (OAM) functionality on RAN elements (via the O1 interface). These open interfaces, supported by standardized software development kits (SDK), facilitate an ecosystem of rApp innovation.
Figure 30: Open programmable network
Data enrichment plays a crucial role in delivering RAN optimization capabilities. True programmability is realized by enhancing decision-making with comprehensive insights based on network data, enabling more precise network management and dynamic performance adjustments. Advanced data management provided by the SMO enables more ubiquitous and low-latent access to dynamically move and control the network data required to realize this programmability.
Driven by efficient data management controls and movement within the network, timely and accurate distribution of diverse data sources will significantly enhance optimization logic and render improved RAN performance. The data can originate within RAN, such as from cell traffic and usage patterns, or from external sources like weather forecasts and IoT device interactions, providing a holistic view for enhanced network optimization. An inclement weather forecast, for example, can trigger allocation of the additional network resources necessary to address the potential higher usage during the forecast window. As a reference, the typical latency for RAN-generated data to reach a self-optimizing network (SON) platform in today’s network is 25–40 minutes. As a result of efficient data management controls and movement, the same data has been made available to an rApp in a few minutes in an AT&T trial of SMO. A key contributor to this efficiency is moving from an aggregated event stream to an event-based data collection and reporting model for RAN telemetry, which took only seconds to reach the SMO’s data collector.
Scope of SMO and programmability via rApps
Programmability introduces a new paradigm in network management through rApps that utilize real-time data for network optimization, network healing, network deployment and network evolution. The capacity of SMO to oversee a larger geographical area than a single NodeB enables consistent and comprehensive network optimizations across broader network zones, enhancing overall service delivery and service level agreement (SLA) management. Creating a “single pane of glass” across SMO instances creates extremely proficient management across the network. All these advancements will enable networks to evolve incrementally rather than via huge network-generation growth spurts. This will also enable service providers to embrace more consistent investment paces for new technology in these areas.
Differentiated connectivity via intent-driven networking
An essential part of network programmability is intent-driven operations which add an abstraction layer above the traditional RAN configuration approach. This allows service providers to specify desired outcomes, or “intents,” which are then autonomously implemented by the network through sophisticated algorithms embedded in the SMO, rApps, RAN and other service and network assurance functions. By aligning network operations with business objectives, network programmability facilitates the creation of differentiated connectivity options that can cater to diverse customer needs, thereby enabling service differentiation at an unprecedented scale that challenges traditional operational models.
Figure 31: Intent-driven networking
In a recent proof-of-concept for intent-based automation, it was demonstrated that an automated solution can meet user requirements for defined minimum or bounded data rates with bounded latency (for example a certain performance level: 1 Mbps throughput with 50 ms latency) for new applications and services for content consumption, generation, or business processes. In this proof-of-concept, Fixed Wireless Access user experience was enhanced and kept within the bounded throughput and latency requirements, even in congested scenarios.
Implementation in AT&T’s network
AT&T is at the forefront of leveraging network programmability to enhance service differentiation, identifying new operating models and rapidly delivering new product opportunities. Through initial proofs-of-concept and trials of intent-driven networks, AT&T is exploring the potential of this technology to provide tangible benefits such as enhanced network flexibility and superior service quality. Furthermore, the deployment of an intelligent automation platform as the foundational architecture underscores AT&T’s commitment to this advanced network paradigm. This intelligent automation platform supports the effective implementation of programmability within the RAN, ensuring that in the near future the network can support differentiated connectivity at scale that is tailored to meet specific customer requirements and evolving market demands.
The journey ahead
AT&T is in the early stages of realizing its programmable network, marking the transformation from the traditional one-size-fits-all, best-effort mobile broadband, toward a new era of performance-based business models built upon differentiated connectivity. A key step being explored is a converged SMO for both core and RAN networks to deliver a unified wireless network management and automation solution, efficiently providing end-to-end visibility across the mobile network. On the business side, they continue to push the boundaries and explore differentiated connectivity across the ecosystem, with proofs-of-concept and live commercial offerings including AT&T Turbo, a premium connectivity offering for consumers to upgrade their experiences.