The rapid expansion of 5G networks brings with it a significant increase in energy consumption, posing a challenge for sustainability. The Aether SMaRT-5G initiative, spearheaded by the Open Networking Foundation (ONF) and funded by the National Telecommunications and Information Administration (NTIA), aims to address this issue. By developing advanced methods for measuring and modeling energy consumption, the initiative seeks to demonstrate and validate energy-saving techniques in collaboration with a diverse vendor/operator ecosystem. Central to these efforts is the creation of the Platform for O-RAN Energy Efficiency Testing (POET).
Energy Consumption Measurement Approaches
Power Distribution Units (PDUs)
Power Distribution Units (PDUs) serve as a vital tool for establishing baseline data on power, current, and voltage supplied to Physical Network Functions (PNFs) and servers. These measurements are taken approximately every 15 seconds, providing a ground truth for power consumption. Once collected, the data is exported to Prometheus/Grafana for detailed analysis. PDUs offer fundamental insights into total energy usage within the network. However, their deployment across extensive networks can be both costly and cumbersome. Thus, while they provide essential foundational data, the challenge remains in balancing their comprehensive measurements against the logistical difficulties of widespread implementation.
As part of the SMaRT-5G initiative, the use of PDUs enables network operators to gather accurate energy consumption statistics, which are crucial for optimizing network performance while minimizing energy wastage. Despite their limitations, the data derived from PDUs informs the development of more refined and targeted energy-saving strategies. It is through these efforts that the initiative aims to promote a more sustainable approach to the rapidly expanding 5G landscape.
Intelligent Platform Management Interface (IPMI)
The Intelligent Platform Management Interface (IPMI) tracks server-reported power and environmental variables by querying the Baseboard Management Controller (BMC). This data is then exported to Prometheus/Grafana, where it can be analyzed in conjunction with PDU measurements. Initial analyses suggest that PDU and IPMI readings align closely, with PDU readings typically registering about 5% higher than IPMI, depending on the load. This close alignment indicates that IPMI can be calibrated to provide a reasonable estimate for PDU readings, making it a valuable proxy for total server energy consumption. In environments where deploying PDUs might be impractical, IPMI offers a more feasible alternative for monitoring power usage.
Nevertheless, it is important to note that IPMI readings are quantized, reporting in steps rather than the continuous readings provided by PDUs. This quantization must be considered in the calibration processes to ensure accurate energy consumption modeling. Despite this limitation, IPMI’s ability to offer close estimations makes it an integral component of the SMaRT-5G initiative’s overall strategy for tracking and optimizing energy consumption across various server setups.
Scaphandre and Kepler
Scaphandre and Kepler are specialized tools aimed at measuring energy consumption in different network environments. Scaphandre leverages the Intel Running Average Power Limit (RAPL) to gauge process utilization and estimate power consumption on bare-metal servers running Control Unit (CU) and Distributed Unit (DU) software. This approach allows Scaphandre to provide detailed insights into the energy demands of these specific network components. By focusing on bare-metal servers, Scaphandre can isolate the power consumption related to the OAI O-RAN system, making it a valuable asset in understanding the intricacies of energy use within these setups.
On the other hand, Kepler is incorporated into the Kubernetes deployment within the testbed. Using an estimation model based on Intel RAPL data, Kepler measures node and container utilization, providing insights into the energy demands of Cloudified Network Functions (CNFs). Although the current models based on CPU utilization show lower energy consumption compared to PDU/IPMI readings due to the exclusion of non-CPU-related energy use, Kepler effectively tracks overall energy consumption trends. Despite its limitations, Kepler’s ability to focus on Kubernetes containers and pods allows for a detailed understanding of the energy demands specific to cloud environments.
Challenges and Opportunities
Granular Insights and Disaggregated RAN Architectures
While PDU data provides a fundamental measurement of total energy usage, it does not deliver the granular insights required to optimize energy management, especially in disaggregated RAN architectures like O-RAN. Understanding energy consumption relative to traffic load by Radio Units (RUs), Control Units (CUs), Distributed Units (DUs), and Core components is critical. The current O-RAN interface capabilities are insufficient for detailed energy consumption measurements, necessitating supplemental approaches such as IPMI, Scaphandre, and Kepler. These tools collectively offer a more nuanced view of energy consumption patterns, allowing network operators to identify specific areas for optimization.
The quest for granular insights is essential in creating more energy-efficient 5G networks. Disaggregated RAN architectures like O-RAN present unique challenges, as the energy consumption across RUs, CUs, DUs, and Core components may vary significantly based on traffic load and other operational factors. By employing a combination of PDUs, IPMI, Scaphandre, and Kepler, the SMaRT-5G initiative aims to develop a comprehensive understanding of these consumption patterns. This understanding forms the basis for targeted energy management strategies, ensuring that resources are allocated efficiently and sustainably across the network’s various components.
Calibration and Modeling Efforts
The calibration of IPMI, Scaphandre, and Kepler readings is essential for accurate energy consumption modeling. Given that IPMI readings are quantized, reporting in steps as opposed to the continuous readings provided by PDUs, this quantization must be accounted for in calibration efforts. Accurate modeling ensures that the data from these tools can be used to develop reliable power consumption Key Performance Indicators (KPIs) and predictive models. Both Kepler and Scaphandre can track total energy consumption trends effectively but require improvements to capture non-CPU-related energy use more accurately.
These calibration and modeling efforts are crucial for the SMaRT-5G initiative to deliver precise and actionable insights into energy consumption. By fine-tuning the readings from IPMI, Scaphandre, and Kepler, researchers can create comprehensive models that reflect real-world usage patterns. This process involves continuous testing and refinement to account for variables like load conditions and hardware differences. The end goal is to develop accurate KPIs and predictive models that network operators can use to enhance energy efficiency across various deployment scenarios, ultimately contributing to the broader objective of sustainable 5G network expansion.
Virtual Network Functions and Load Impact
Separate Estimation of Power Consumption
To evaluate energy consumption across various virtual network functions (VNFs) such as Core, CU, and DU, TCP tests were conducted with different user equipment (UE) loads. Kepler can separately estimate power consumption for 5G Core, O-CU, and O-DU in Kubernetes deployments, providing valuable insights into the energy demands of these individual components. This separate estimation is crucial for understanding the specific energy needs of each VNF, allowing for targeted optimization efforts. By isolating the power consumption of Core, CU, and DU functions, network operators can make more informed decisions about resource allocation and energy-saving techniques.
Understanding the distinct energy consumption patterns of different network components is vital for developing effective energy-saving strategies. With the ability to separately estimate power consumption for Core, CU, and DU functionalities, network operators can identify which components are the most energy-intensive and prioritize them for optimization. This approach helps to create a more detailed and precise energy management framework, ensuring that each part of the network operates as efficiently as possible. The insights gained from these separate estimations inform the broader strategies for achieving sustainable and energy-efficient 5G networks.
Load Impact on Energy Consumption
The energy consumption of the Distributed Unit (DU) increases most significantly with load changes compared to the Core and Control Unit (CU). This observation highlights the importance of focusing on the DU for energy optimization efforts. By understanding the load impact on energy consumption, network operators can implement targeted energy-saving techniques to improve overall network efficiency. The DU’s sensitivity to load variations makes it a priority area for optimization, as even small improvements in DU energy management can have a substantial impact on the network’s overall energy use.
Load impact studies provide critical insights into how different components of the network respond to varying traffic conditions. The Distributed Unit (DU) has emerged as a key focus area due to its significant energy consumption changes in response to load variations. By prioritizing the optimization of DU energy use, network operators can achieve substantial energy savings and enhance the overall efficiency of the network. These studies underscore the need for continuous monitoring and adaptive energy management strategies that respond dynamically to changing load conditions, ensuring that 5G networks remain both efficient and sustainable.
Linking Energy Consumption with Network Performance
End-to-End Measurements
A crucial goal of the SMaRT-5G initiative is to link energy consumption with network performance, deriving key performance indicators (KPIs). End-to-end measurements involve analyzing data volumes and throughput across the network. These measurements provide insights into the relationship between energy consumption and network performance, helping to identify areas for improvement. By understanding how energy use correlates with network performance metrics, operators can develop strategies that optimize both energy efficiency and service quality.
End-to-end measurements are a vital component of the SMaRT-5G initiative’s efforts to integrate energy consumption data with performance metrics. These measurements allow researchers to capture a holistic view of network operations, identifying how energy use impacts data throughput and overall efficiency. By linking these aspects, the initiative can develop more robust KPIs that reflect the true performance of the network. This comprehensive approach ensures that energy optimization efforts do not come at the expense of network quality, striking a balance between sustainability and service excellence.
O1 and E2-Based KPIs
Utilizing the OAI DU O1 solution for uplink/downlink throughput metrics and downlink PRB load is another approach to linking energy consumption with network performance. The initiative also leverages several KPIs available over the E2 interface, such as E2-SM KPM. These KPIs are essential for developing a comprehensive understanding of the energy-performance relationship and guiding energy. By using these advanced measurement techniques, the SMaRT-5G initiative aims to build a more detailed and accurate picture of network operations.
Leveraging O1 and E2-based KPIs offers a more granular view of the network’s performance in relation to its energy consumption. The use of OAI DU O1 solutions and E2 interface metrics enables the initiative to capture detailed data on uplink/downlink throughput and other key performance metrics. This data is invaluable for developing precise KPIs that align closely with energy consumption patterns. By integrating these KPIs into their analysis, the SMaRT-5G initiative can develop targeted strategies for minimizing energy use while maintaining high-performance standards across the network.
Conclusion
The swift expansion of 5G networks leads to a notable surge in energy consumption, presenting a challenge for sustainability. Addressing this issue, the Aether SMaRT-5G initiative, led by the Open Networking Foundation (ONF) and funded by the National Telecommunications and Information Administration (NTIA), aims to tackle the energy demands of growing 5G infrastructure. This initiative focuses on developing sophisticated methods to measure and model energy use, striving to prove and validate energy-saving maneuvers in partnership with a varied group of vendors and operators. Central to these efforts is the creation of the Platform for O-RAN Energy Efficiency Testing (POET), which plays a crucial role in testing and enhancing energy efficiency in real-world scenarios. By leveraging collaborative efforts and advanced technology, the Aether SMaRT-5G initiative seeks to make 5G networks more sustainable and energy-efficient, ensuring the benefits of 5G can be enjoyed without compromising environmental responsibility.
