The open radio access network (ORAN) is a new industry standard for RAN that establishes interfaces to promote interoperability among equipment from different vendors and offer network flexibility at a lower cost. It leverages the benefits of network softwarisation and artificial intelligence (AI) to enhance the operation of RAN devices. However, building a brand-new ORAN architecture from the ground up is difficult. There are several challenges to consider while setting up and testing ORAN networks.
A look at the key challenges associated with ORAN deployment…
The ORAN ecosystem presents a wide range of vendors and RAN components, enabling operators to select and combine solutions from different suppliers. This flexibility and choice come with the challenge of ensuring interoperability among different components. When an issue arises in the network, it can be difficult to identify and isolate the problem due to the complexity of the environment.
The different implementations of ORAN solutions by different vendors may lead to compatibility issues and potential conflicts between components. Moreover, the complexity of the environment may make it difficult to identify the root cause of an issue, and vendors may be hesitant to take responsibility for problems that are not their fault. Therefore, ensuring seamless interoperability is essential for the success of ORAN deployments.
The efficient allocation and utilisation of network resources is a significant challenge for ORAN systems. In the current cellular network, the effective management of heterogeneous traffic flow in relation to network capacity remains a key issue. ORAN networks must be able to adapt to the dynamic resource requirements of network services and users (such as network slicing). The ever-increasing network traffic also underscores the need for intelligent and automated solutions.
The introduction of RAN intelligent controllers (RICs) has facilitated the deployment of intelligent solutions in the form of xApps and rApps to manage complex network activities. These solutions rely on data from the lower network level, such as distributed units (DUs), for training and accurate prediction. However, the data generated at DUs contains user-sensitive information, which could be exposed when shared with RICs. Furthermore, the multi-vendor ecosystem of ORAN may create confidentiality challenges when handling such user-sensitive data.
The lack of an end-to-end testbed for the ORAN system poses a significant challenge to the implementation and testing of various use cases. The ORAN Software Community offers individual code bases for ORAN functionalities and interfaces (RICs and service management and orchestration), but there are no clear guidelines for integrating these functionalities with existing implementations from other open-source communities such as OpenAirInterface and Magma core.
Meanwhile, a number of initiatives are being undertaken to address the issue of testbed availability in ORAN. However, despite these efforts, testbed availability in ORAN is likely to remain a challenge for the foreseeable future.
The integration of machine learning (ML) and AI into the ORAN concept plays a critical role in supporting the next generation of wireless networks. However, the adoption of ML and AI also introduces new security vulnerabilities that must be addressed.
One of the most common threats is data poisoning attacks, wherein a hacker intentionally corrupts the data used to train, test, or validate AI and ML models. This can be done by injecting malicious data into the training dataset or by modifying existing data. The hacker can then gain access to the data by infiltrating the fronthaul, midhaul, xApps, or rApps.
Data poisoning attacks can have a significant impact on the performance of AI/ML models, leading to incorrect predictions, decision-making and classification. They can also be used to disrupt the operation of the ORAN system through denial-of-service attacks or other malicious activities.
The management of data is fundamental to all AI and ML activities in the ORAN system. A major challenge lies in storing, categorising and selecting RAN-related data aligned with specific use-case requirements. Another challenge lies in accessing proprietary data owned by base station owners, as sharing such data is hindered by privacy and competition concerns. A lack of open-source data for research purposes also poses challenges.
The use of intelligence within ORAN helps address the ever-growing complexity of mobile networks driven by the increasing demand for user data. This is achieved by embedding intelligence at the network and component levels to manage and optimise resources. However, a major challenge for the ORAN system is the orchestration of AI/ML models. There are still concerns regarding the selection of the right AI/ML model, appropriate deployment locations and resources, and the timeframe for making inputs available. Another major challenge in intelligence management is handling fault prediction (particularly drifting) in AI/ML results.
The traditional design of cellular networks prioritises high-quality network performance over energy conservation. As a result, RANs consume more than half of their total energy. The disaggregation of network functions in ORAN into various hardware components contributes to increased energy consumption. Energy efficiency has become a major concern for network operators, as the amount of energy consumed is directly proportional to the operational costs of the network.
The disaggregation of RAN functionality can provide several technological and performance benefits, but it also introduces the challenge of system unavailability. The introduction of open markets and interfaces, as well as the virtualisation of RAN functions, raises hardware and software security concerns. The location of RAN functionalities has a significant impact on availability in an ORAN system. Further, the performance of RIC applications, such as xApps and rApps, must be resistant to outages to ensure system availability.
The road ahead
Open RAN has the potential to deliver transformative value across the telecom ecosystem, but its rapid adoption will depend on addressing concerns about interoperability, testbed availability, energy consumption, data management, confidentiality, and more. The transition to Open RAN is expected to be a phased process, with deployments accelerating as all the challenges are addressed in a timely manner.