The innovation potential of Non Real-time RAN Intelligent Controller
Senior Architect
Senior Architect
Senior Architect
The value of intent-based management of RAN
As the mobile networks evolve from 4G to 5G, service providers face a new reality of requirements and expectations at the same time as new technologies can help them mitigate the effects of increased complexity. In a typical radio-access network there are literally millions of decisions taken every second about which user to serve over the radio interface and how. Each of these decisions contribute to the service quality and the prioritization among users and services in case of conflicts.
Traditionally these micro decisions are governed by a combination of supplier design choices and network configuration parameter settings done by the service provider. In the relatively simple 2G systems, the effect of a configuration change was mostly possible to understand. In today’s more sophisticated multi-service 5G networks it is virtually impossible, in a cost-efficient manner, to predict the effect a given set of configuration changes will have on the end-user services.
However, the intent of the RAN remains the same, to offer connectivity to the service providers’ customers in a profitable way. The idea of intent-based management for RAN is to evolve the configuration of the RAN from setting technical parameters (such as hand-over thresholds) and instead allowing service providers to specify the connectivity service itself, prioritizing across users and services, based on business intent. Non Real-Time RAN Intelligent Controller (Non-RT RIC) is a concept developed by O-RAN Alliance to realize Intent-based management, built on principles of automation and AI and Machine learning. The Non-RT RIC brings genuinely novel capabilities to the system and addresses use cases that were previously out of reach, with the ability to set policies per user and data enrichment information for RAN optimization.
Intent-based management based on Non-RT RIC can be applied to Cloud RAN to enable a high degree of network programmability and can equally well be applied to purpose-built RAN to enable a wide variety of automation and optimization use-cases that are not possible today.
Deep dive on the non-RT RIC technology
The Non-RT RIC is part of the Service Management and Orchestration system (SMO) as defined by the O-RAN Alliance and consists of a platform plus a set of microservices (named rAPPs by O-RAN Alliance) that represent the intelligence (see figure 1). The design of the system is based on these principles:
- Access to information: There is a wealth of contextual information – not available in the RAN – with potential to improve the radio-resource management and RAN performance in general. This includes application-level information, cross-domain information, UE positions and mobility trajectories as well as external information.
- Dynamic optimization: Traditionally management and orchestration have been done on the timescale of hours. With automation and improved interfaces, the Non-RT RIC can optimize the RAN on a time scale down to half a second.
- User level service assurance: Optimizing the RAN on a user level (in addition to per-node level) enables the Non-RT RIC to address a wide set of use-cases that were previously out of reach.
- AI/ML over engineered programs: The intelligence in RAN control is gradually moving to AI/ML-based software, and the Non-RT RIC is designed for AI/ML from day one.
- Innovation for openness: It is possible to build an open eco-system of intelligent controller software where applications (rAPPs) feed each other with data and insights
Figure 1 - The Non-RT RIC uses the interfaces A1, O1 and O2 to interact with the RAN on timescales above 500ms
The placement of the Non-RT RIC in the SMO and not in the RAN is to secure access to contextual data and use it to optimize the RAN, something that the RAN nodes CU, DU and Near RT-RIC can’t do. One example is the coordinated optimization of radio and transport to adjust user-level throughput based on application performance and contextual prioritization, which can’t be achieved in existing networks.
Near-RT RIC is a new element, defined in O-RAN, inside the RAN aiming for radio resource management. Currently, it is unclear what performance benefits Near-RT RIC can provide beyond that of a 3GPP technology.
Shorter time scales improve network control and user experience
The Non-RT RIC introduces a new control loop in the system on a time scale between traditional management (minutes/hours) and the RAN-internal control loops (between 50 microseconds and 100 milliseconds). Setting interacting control loops on different time scales is a proven design pattern to maintain system stability and separation of concern between the loops so by setting the Non-RT-RIC time scale to around half a second we avoid potential conflicts and instabilities that can occur. As a contrast, the CU-CP and the Near-RT RIC both act on the same 100 ms time scale which is a potential source of conflicts Non-RT RIC adds the ability to send optimization policies enrichment information targeting an individual user by adding the new A1 interface that complements the per-node configuration interface O1.
Combining user-level granularity, shorter control loop time scale and the access to outside-RAN information, the Non-RT RIC brings novel system capabilities that were not available in any earlier generations of mobile systems. Adding rapid development of AI/ML technologies to the mix leads us to predict that the Non-RT RIC will drive significant innovation and address use-cases that were previously not possible to solve and that we will look into in the next section.
To capitalize on this, the Non-RT RIC is designed with openness and innovation in mind and follows a micro-service architecture where intelligent applications (the “rAPPs”) are deployed on top of the Non-RT RIC platform. A standardized information model and the open interfaces A1, O1 and O2 makes it possible to have rAPPs from both incumbents, third parties and service providers alike, that together, can steer any RAN.
rAPPs are integrated to the rest of the system through what is called the R1 interface, presently technically studied in O-RAN. The R1 interface will support easy onboarding of rAPPs and will give service providers controlled authorization so that rAPPs can be onboarded at low cost, trialed in limited parts of the network, evaluated and then either scaled out to a larger part of the network (if successful) or fail early, pulled out and re-designed (if not so successful).
Categories of use-cases enabled
The non-RT RIC enables a wide variety of new use-cases. Some of these use-cases can be seen as enhancement of existing functions, while other are entirely new. On a high-level, the non-RT RIC use-cases can be categorized as follows:
- User-level policies: The Non-RT RIC can influence the RAN behavior by means of declarative policies, per user and per user group. This enables optimized treatment of individual users based on their special requirements. In the next section we will explore this in detail.
- RAN enrichment data: The Non-RT RIC can provide additional information to the RAN, for better resource optimization. For example, the Non-RT RIC can provide information about weather conditions to the RAN so that the it can change behavior, such as prioritizing frequencies that are less affected by rainy conditions.
- Enhanced self-organizing-network (SON) functions: Using fast closed-loop automation the Non-RT RIC can enable more advanced SON use-cases. For example, the dynamic orchestration of radio and transport together can provide more efficient end-to-end resource usage.
Service assurance for first responders – a real-life application of Non-RT RIC
Imagine that a large fire is ongoing in an industrial area. First responders have arrived on-site, bravely fighting the fire. Each first responder streams a live video, creating situational awareness for the task force leader. The live feed of some first responders are more important than others – the closer they are to the fire, the more important information.
Non-RT RIC provides the ability to prioritize radio resources all the way down to the individual first responder based on contextual, external information such as the location of the fire (e.g. captured through location sensors in the building) and the location of the first-responder (e.g. captured through GPS). In existing networks this is not possible. Today, it is only possible to perform service assurance per group of users (i.e. all the first responders) or per node (i.e. everyone close to the fire).
With Non-RT RIC, an rAPP can be designed to provide this kind of granular service assurance, improving the quality-of-experience where it is really needed and, in this case, potentially taking out the fire faster. It also opens up completely new doors for innovation and business opportunities for service providers.
Figure 2
Ericsson is very excited to see the industry move towards an architecture that enables an intent based management of the Radio Access Network, AI-enabled closed-loop automation and end-to-end optimization. We believe this will bring industry value by enabling new use-cases and improving end-user performance. Service providers can benefit from this innovation no matter the type of RAN infrastructure - purpose-built, cloud native or open. Ericsson stand ready to guide and collaborate with our customers as they embark on this journey.
Watch the Innovation potential of Non Real-time RAN Intelligent Controller video
