As Open RAN gains momentum, so does the need to understand how vendor diversity and network disaggregation impact both performance and energy efficiency. Two NTIA PWSCIF NOFO1 projects, ACCoRD and SMaRT-5G, are tackling these challenges from different but complementary angles. This emerging collaboration could help push the industry forward with better tools, methods, and insights.
ACCoRD (Acceleration of Compatibility and Commercialization for Open RAN Deployments): Real-World Testing of Disaggregated RAN
The ACCoRD project focuses on Testing & Evaluation (T&E) in 5G Open RAN. Its core mission is to validate interoperability and performance when mixing and matching network function vendors, e.g., O-RUs from one vendor, and O-DUs/CUs from another. These are not just theoretical pairings because ACCoRD testing takes place within real testbeds that are directed by a consortium of Mobile Network Operators (MNOs) to ensure that test results obtained from evolving commercial products are reproducible and on the path to deployment-readiness.
Funded by NTIA PWSCIF, ACCoRD’s ultimate goal is to support the availability, affordability, and accessibility of testing and evaluation in the United States in support of the development, deployment, and adoption of open and interoperable, standards-based radio access networks. Up until recently, the ACCoRD testbeds’ primary focus has been on KPIs (Key Performance Indicators) such as throughput, latency, and stability, but not yet on energy consumption; nevertheless, the infrastructure, test scenarios, and tools already in place provide a valuable foundation for energy-aware experimentation.
SMaRT-5G ( Sustainable Mobile and RAN Transformation 5G)
SMaRT-5G is a collaborative effort to develop and enable ML-driven energy savings solutions for mobile networks, led by Open Networking Foundation (ONF)/Aether Projects – Linux Foundation. A major SMaRT-5G initiative is the 5G energy measurements, metrics and modeling R&D project, which is being carried out collaboratively between ONF/Aether and Rutgers University/WINLAB, funded by NTIA, with significant support from Keysight Technologies, through advanced test solutions and domain expertise.
The SMaRT-5G project brings a complementary focus: understanding how 5G networks consume energy, and how that energy utilization evolves with load, user count, and service type. The answers to these questions are based on both physical measurements and modeling of energy efficiency, for both end-to-end O-RAN systems as well as emulator-based component-level testing. To gather trustworthy energy consumption data, both Dynamic DC Power Supply and industrial monitored Power Distribution Units (PDUs) serve as the primary “ground truth” for physical network functions (PNFs) like O-RUs, physical O-DUs and infrastructure servers. Energy and performance KPIs are gathered using telemetry network methods which are aligned to existing, or anticipated, O-RAN interface specifications: (1) O-RUs may report power and other metrics via the Open Fronthaul M-Plane interface; (2) O-DUs/O-CUs may report power and performance metrics through the O1 interface. For virtualized/cloudified network functions (VNFs/CNFs), there are open-source tools, such as Kepler, Scaphandre, Redfish and ipmitool, that are being used to monitor both container- or pod-level as well as bare-metal power consumption, and these may be reported through the O2 interface. All data are pushed into Prometheus and visualized through custom Grafana dashboards. Energy usage is tracked at various levels: (1) per PNF (entire O-RU or physical O-DU); (2) per container or pod (in VNF/CNF-based setups for DU, CU, 5G Core).
The goal is to derive energy consumption models that typically start with empirical approaches, i.e., based on measured data, while keeping analytical and simulation-based options on the table. Key inputs for modeling include baseline/idle power consumption for all components as well as total RF power, number of UEs (real or emulated), data throughput and volume, latency, and the resource utilization (e.g., number of Physical Resource Blocks (PRBs) in use). KPIs under exploration include watts per PRB utilized, watts per throughput unit, and watts per system load percentage. These are still exploratory, but the goal is to standardize the most useful KPIs across testbeds once they prove reliable.
To evaluate energy behavior across different loads and usage patterns, iperf is used for synthetic traffic generation, and UE Emulators create more realistic user scenarios and dynamic traffic patterns including bursty traffic, full-buffer streaming, and idle phases. Test scenarios are designed to cover idle mode, low-load conditions, and peak-load performance.
Where do these two NTIA PWSCIF NOFO1 projects intersect?
Infrastructure Compatibility: ACCoRD’s diverse and realistic testbeds, with a range of O-RU and O-DU/CU combinations, offer exactly the kind of environments that SMaRT-5G needs to test its energy consumption models. With the addition of monitored PDUs, these testbeds could deliver the “ground truth” energy data needed.
Methodology Feeding Into Test Cases: ACCoRD is beginning to define test cases that include energy measurement and efficiency aspects. Here, SMaRT-5G’s experience in designing energy-relevant scenarios and KPIs is extremely valuable. The two projects can collaborate to ensure the test case definitions in ACCoRD are informed by sound methodology and aligned with SMaRT-5G’s modeling needs.
Shared Testbeds and Network Functions: The ACCoRD testbed at Rutgers University WINLAB is already shared between both projects, offering the perfect opportunity for joint experiments. The network functions themselves could also be re-used, providing consistent baselines for energy and performance analysis.
Tooling Reuse Potential: While SMaRT-5G’s tooling is not yet fully portable (collection scripts are tailored per testbed), some dashboards and Prometheus-based visualization tools could be reused in ACCoRD testbeds, once data collection is adapted.
What’s Next?
This collaboration is still evolving, but one clear next step is a joint effort to define energy-impact test cases. By combining ACCoRD’s practical testing experience with SMaRT-5G’s modeling and methodology insights, the projects can help the industry answer key questions. How much does vendor diversity impact energy use? Can energy consumption be accurately predicted from performance metrics, load and baseline characterization? What’s the best way to measure energy efficiency in disaggregated RAN deployments?