Using dApps for real-time control and management
The Open Radio Access Network (Open RAN) ecosystem is undergoing a radical transformation driven by the need for more efficient, flexible, and intelligent network management. Central to this evolution are the RAN Intelligent Controller (RIC) applications, including rApps, xApps, and the newly emerging dApps. These applications serve distinct roles in optimizing network performance, with dApps standing out for their capability to handle real-time control loops below 10 milliseconds, offering unparalleled control and management.
Understanding the landscape: rApps, xApps, and dApps
rApps: Strategic long-term management (Non-Real Time RIC)
rApps run on the Non-Real-Time (Non-RT) RIC and they are responsible for long-term control actions and decision-making that requires timescales greater than one second. These applications are essential for strategic network management and optimization tasks that do not demand immediate responsiveness. By managing longer-term network operations, rApps ensure sustained performance improvements and strategic alignment with business goals.
xApps: Near-Real-Time Control
Operating on the Near-Real-Time (Near-RT) RIC, xApps handle control loops ranging from 10 milliseconds to a few seconds. They enable closed-loop control for use cases such as traffic steering, Handover (HO) management for Vehicle-to-Everything (V2X), and fast loop RAN slice SLA optimization, which offers significant improvements in network responsiveness and efficiency. xApps bridge the gap between immediate operational needs and long-term strategies to provide a balanced approach to network optimization.
dApps: Real-Time precision
Distributed applications (dApps) run directly on RAN elements, such as Central Units (CUs) and Distributed Units (DUs), to manage real-time control loops below 10 milliseconds. dApps focus on critical radio management tasks like resource scheduling, beamforming, and modulation. By operating closer to the data source, dApps ensure fine-grained management and reduced overhead, which offers faster control for these essential functions.
The emergence and need for dApps
Openness and programmability in the RAN bring a radical transformation to the cellular ecosystem through RAN Intelligent Controllers (RICs). The O-RAN RICs can improve performance in various use cases, including (but not limited to) traffic steering, load balancing, slicing, and energy efficiency. These use cases are supported by closed-loop control running at timescales of 10 milliseconds to 1 second (Near-Real-Time RIC with xApps) and of 1 second or more (Non-RT RIC with rApps).
However, current specifications do not provide clear mechanisms, procedures, and architectures to execute real-time control loops operating at timescales below 10 milliseconds that are not hardware-based and embedded in the RAN components. Implementing sub-10 millisecond control loops involves overcoming limitations in current hardware processing and software algorithms that can respond within such a narrow timeframe, as well as in the programmability and interfaces of the systems where such control loops need to be embedded.
The concept of dApps, which emerged in 2021 -22, introduces distributed applications that enhance and extend the capabilities of existing xApps and rApps. These applications enable operators to implement fine-grained, data-driven management and control in real time at the O-RAN Central Unit User Plane (O-CU-UP), Control Plane (O-CU-CP), and Distributed Units (O-DUs). Researchers have successfully implemented dApps and related real-time RAN control features, demonstrating their effectiveness in making real-time decisions on spectrum sharing, scheduling, RAN slicing, and policy enforcement, driven by AI and ML solutions.
Beyond network performance and optimization, dApps are also applicable to time-sensitive commercial use cases such as industrial automation, AR/VR, financial trading, and gaming, where ultra-low latency and real-time processing are critical
Benefits of dApps in Open RAN
dApps address two critical limitations of the current architecture simultaneously – the lack of control loops within a period faster than 10 milliseconds and the lack of interaction and programmability on the user plane. The main benefits that dApps bring to the O-RAN architecture include:
- Real-time interactions with the RAN protocol stack: The ability to execute intelligence in a unit co-located with O-CUs and O-DUs opens up many new customizable and programmable inference and control loop capabilities, including beam management, scheduling profile selection, packet tagging, dynamic spectrum access, and Quality of Service (QoS) enforcement.
- Inference based on user plane data: The O-RAN architecture has primarily focused on the control plane of the network. Embedding programmable components within the user plane offers significant benefits, such as anomaly detection, spectrum sensing, fingerprinting, and beam management. However, it also presents challenges in avoiding compromises to network performance while ensuring security, privacy, and timing/bandwidth constraints.
Real-world applications
Beam management
dApps can implement AI-driven beam selection logic to enhance spectrum utilization and reduce interference.
Spectrum sensing
This enables spectrum sensing applications like waveform classification and interference management, which includes:
- Interference detection
- Interference classification and characterization
- Interference mitigation
- Spectrum hole detection
- Incumbent detection
Once interference is classified and characterized, proper mitigation techniques (such as transmit power adjustment, PRB blanking, carrier aggregation, beamforming, and beam muting) can be deployed in real-time. dApps hosting spectrum sensing algorithms can achieve real-time spectral awareness directly at O-DUs without interacting with external systems like the Environmental Sensing Capability (ESC) used in Citizens Broadband Radio Service (CBRS). Other applications include automated Bandwidth Parts (BWP) reconfiguration to avoid interference at certain subcarriers and numerology and bandwidth reconfiguration to fit transmissions within spectrum holes.
Real-Time scheduling
Scheduling is a key activity enforced in the O-DU, involving several network components up to the core network. Since scheduling requires real-time decisions at the Transmission Time Interval (TTI) level, controlling scheduling decisions is not compatible with closed-loop control orchestrated by xApps or rApps. Here, we discuss two potential applications of dApps for the improvement of the scheduling function.
Ultra-Low Latency applications
By managing scheduling and slicing at sub-10 millisecond timescales, dApps can support Ultra-Reliable Low-Latency Communication (URLLC), which are crucial for applications like autonomous driving and industrial automation.
Traffic analysis, interference detection, and mitigation
dApps can be instrumental in analyzing traffic patterns, detecting interference, and implementing mitigation strategies to ensure seamless network performance.
LLM-based CSI feedback, channel estimation, and coding
Leveraging advanced AI techniques, dApps can provide real-time feedback on Channel State Information (CSI) to enhance channel estimation and coding processes.
Challenges and architectural evolution
While promising, the implementation of dApps presents several challenges:
- Resource management: With multiple dApps sharing the same physical resources at DUs/CUs, intelligent resource allocation is crucial to avoid conflicts and ensure efficient operation.
- Standardized interfaces: Defining and standardizing the interfaces connecting dApps to CUs/DUs is essential for seamless data collection and control.
- Orchestration: An orchestration framework is necessary to decide which tasks are better for xApps compared to dApps, and then, to mitigate any conflicts between them.
Proposed architecture for dApps
A robust architecture for dApps involves:
- dApps as container architecture: Utilizing containerization to ensure flexibility and scalability.
- Southbound and Northbound interfaces: Establishing standardized interfaces for effective communication and data exchange.
- Conflict mitigation: Implementing strategies to manage and resolve conflicts between dApps and other network elements.
- Intent-based orchestration: Using intent-based policies to guide the orchestration and deployment of dApps, ensuring alignment with overall network objectives.
Conclusion
dApps represent a significant advancement in the evolution of RAN management, bridging the gap between immediate operational needs and strategic network goals. As the industry continues to embrace these innovations, the potential for transformative impacts on network performance and user experience grows exponentially.
