Telecom operators all over the world are warming up to the idea of replacing their legacy radio access networks (RANs) with Open RANs (O-RAN), as th­ey look to lower their capex and opex ami­dst rising capital intensity and subdued subs­criber and revenue growth. By disaggregating the software and hardware elements of networks, O-RAN has the potential to reduce operators’ dependency on a single vendor and bring down network-related costs. Further, by enabling operators to increase the number of their network infrastructure partners, O-RAN can help operators build better and more cost-effective 5G networks.

Although this technology is still at a nas­cent stage and is yet to achieve mass com­mercialisation, recent advances are expected to give a major boost to its adoption. Operators are now leveraging next-generation technologies such as artificial intelligence (AI), machine learning (ML), virtualisation and big data analytics to automate the software components of the O-RAN architecture. Besides enhancing ope­ratio­n­al efficiency, the use of these tech­nologi­es is helping operators make their networ­ks more intelligent and scalable.

Making RAN intelligent using AI and ML

Operators have started harnessing ad­vanc­ed AI and ML applications to im­prove the network performance of O-RAN and en­hance end-user experience. O-RAN gives much greater flexibility to integrate AI and ML algorithms into the RAN management. The AI algorithms help operators analyse vast amounts of tra­ffic data and network load in real time, without impacting the capacity of the network. This makes it possible to obtain instant traffic predictions in the network, without the need for extra hardware or manual site visits. AI and ML are currently being used for forecasting quality parameters, detecting anomalies in the system and predicting failures. Further, with the help of AI and ML, operators are able to adjust network conditions, thereby ensuring proper load balancing and seamless transfer of an active call or data session from one cell in a cellular network or from one channel to another.

The recent introduction of a RAN intelligent controller (RIC) is facilitating the use of AI and ML tools in the O-RAN. RIC is a new optional virtualised 5G optimisation technology that adds programmability to an existing or new RAN and allows self-optimising networks (SON)-like capabilities to be added. Using RIC, operators are able to operate applications such as mobility management, admission control and interference management in the O-RAN framework. RIC provides advanced control functionality, which deli­vers increased efficiency and better radio resource management. These control functionalities leverage analytics and data-driven approaches, including advanc­ed AI and ML tools to improve resource management capabilities.

At present, trials are under way to test and demonstrate AI and ML capabilities in the O-RAN space. In early 2021, Nokia and China Mobile completed live trials of AI for RAN applications on the operators’ 4G and 5G networks. The two companies tested an AI-based real-time user equipment (UE) traffic bandwidth forecast in Shanghai and automated network anomaly detection in Taiyuan. In addition, a RAN Intelligence Controller was deployed in edge cloud infrastructure. Further, in Shanghai, the trial confirmed that AI-based real-time UE traffic prediction ac­curacy exceeded 90 per cent in a live 5G network test.

Scaling networks through cloud

The decoupling of the software elements of the network from hardware in the O-RAN framework is making it easy for operators to migrate to a cloud-native network architecture. In a cloud-native syst­em, ap­plications can be developed and operated in virtual cloud environments. Fur­ther, applications in such a system can be carved up into smaller units called microservices and a group of smaller and interconnected microservices can then be used to replace large applications of a traditional RAN. The cloud-native model enables workflow orchestration and network automation that facilitates easy de­ployment of applications, scaling up of systems and repair of network faults. This is extremely beneficial for remote locations that are otherwise difficult to be served wi­th­out manual intervention. Seve­ral major cloud vendors such as Amazon, Goo­gle and Microsoft have partnered with telecom operators globally to facilitate the migration to cloud-native O-RANs. Ind­us­try ex­perts believe that the involvement of big public cloud providers is likely to provide a huge impetus to O-RAN adoption.

In fact, telecom operators have also started leveraging recent advancements in the cloud technology such as edge compu­ting to improve their O-RAN experience. Edge computing is a distributed com­puting framework that brings computation and data storage closer to the data so­urces. In an O-RAN framework, edge computing can help operators virtualise their networks and run internal operations more efficiently.

Using big data analytics for network insights

With the use of big data analytics in an O-RAN system, operators can obtain a visual representation of patterns and abnormalities in their networks and enable network intelligence across all the O-RAN network elements, including the O-RAN cont­ro­ller, edge core and security gateway pro­duct suites. The technology can be used to store, process and analyse a large amount of complex data, both structured as well as unstructured and the insights from the same could then be used to enhance the service quality for retail end-users and enterprises. Several operators have already started using big data solutions to ensure fast visibility and analysis to accelerate network optimisation and fault resolution. By obtaining massive, dynamic network maps and multidimensional analytics that update in seconds, operators are able to achieve in-depth multilayer troubleshoo­ting from the core to the cell site and connected devices.

Making networks versatile with virtualisation

Virtualisation has emerged as a significant part of the O-RAN architecture roadmap. By virtualising their RAN frameworks, operators can accelerate service innovation and introduce intelligence in RAN control. This not only enables interoperability am­ong RAN components from different sour­ces, but also improves the supply chain security. In addition, introducing virtualisation in O-RAN reduces network capex and opex costs. Operators have started virtualising their O-RAN systems by leveraging software-defined networking and network function virtualisation techniques.

Conclusion

O-RAN architecture offers several ad­vantages over the traditional RAN, key among them being the possibility to apply and combine both, non- and near-real-time analytics, ML and decision-making tools to analyse the various elements of a network. The adoption of O-RAN is ex­pected to soar significantly in the next few years, driven primarily by the roll-out of 5G services and the growing need for operators to undertake network deployments and upgrades at lower costs. Ac­c­ording to industry estimates, O-RAN revenues will account for more than 10 per cent of the overall RAN market by 2025.

The addition of next-generation technologies could further fuel the uptake of O-RAN by telecom service providers. By enabling the separation of control functions from the hardware fabric, O-RAN facilita­tes the introduction of standardised control interfaces, which can be embedded with solutions, providing real-time network analytics. O-RAN, therefore, coupled with next-generation solutions, can help in developing autonomous, and self-optimising networks, which can be scaled up easily, depending on operators’ requirements and fixed without manual intervention. Going forward, the use of advanced AI, ML, big data analytics and cloud-native automation tools is likely to witness a significant uptick as operators look to enhance their O-RAN experience by obtaining intelligent insights related to the management and operation of their networks.

The addition of new solutions is, however, likely to be challenging and wou­ld entail disruptive changes to many existing workflow stages and algorithm designs and require re-engineering in a staged ap­p­roach. The potential benefits of these te­ch­nologies, however, outweigh the costs associated with the redesigning of net­works and the latter would decrease further as the O-RAN architecture achie­ves mass commercialisation.