Open RAN Handbook 2nd Edition

Open RAN Handbook

2ndEdition: a Guide for
Sustainable Networks

C2 General
 Open RAN Handbook 2nd Edition

Authors:
Vodafone Group PLC and Keysight Technologies

Version:
1.0

Document Status:
Released

C2 General
Open RAN Handbook 2nd Edition

© 2025 Vodafone Group PLC

All rights reserved.
This document comprises proprietary and confidential
information and copyright material belonging to Vodafone
Group PLC. It must not be reproduced, used, published, or
disclosed to third parties without the prior written consent
of Vodafone Group PLC.
The information in this document is subject to change
without notice.
All trademarks acknowledged.
Vodafone Group PLC
Vodafone House
The Connection
Newbury, Berkshire
RG14 2FN
England

3
Version 1.0
Open RAN Handbook 2nd Edition
Document Control

Revision History

Version Date Authors Change Summary

1.0 Feb. 2025 Akash Dumbre (Vodafone) First release
Eng Wei Koo (Keysight)
Christian Nunez-Alvarez (Keysight)
Jasper Li (Keysight)

Key Vodafone Reviewers

Department Role Name

Technology Lab Test & System Authority Akash Dumbre
Radio Hardware Development Open-RAN
Technology Miguel Angel Flores
Engineer

Vodafone Approval

Department Role Name

Technology Lab Test & System Authority Akash Dumbre

Technology OpenRAN Senior Manager Dan Shannon

Technology OpenRAN Product Integration Manager James Grayling

Keysight Approval

Department Role Name

Wireless Open RAN Director Eng Wei Koo

4
Version 1.0
Open RAN Handbook 2nd Edition
Contents

Document Control .................................................................................................4

Revision History ......................................................................................... 4
Key Vodafone Reviewers ........................................................................... 4
Vodafone Approval.................................................................................... 4
Keysight Approval ...................................................................................... 4

Contents ...............................................................................................................5

1 Executive Summary........................................................................9

2 Motivation and Challenges ........................................................... 10

3 Standardization and Industry Forums ........................................... 11

3.1 3GPP ............................................................................................... 11
3.2 O-RAN ALLIANCE ............................................................................ 13

4 Open RAN Architecture and Components ..................................... 15

5 Open RAN Energy / Performance Benchmarking Considerations ... 17

6 Open RAN Energy Benchmarking Test Scenarios, Methodology and

Setups ......................................................................................... 19
6.1 O-RU ............................................................................................... 19
6.1.1 Vodafone perspective.......................................................... 19
6.1.2 O-RAN specifications ........................................................... 20
6.1.3 Different options ................................................................. 23
6.1.4 Collaboration test setup ...................................................... 24
6.1.5 Findings and results ............................................................. 27
6.2 O-DU / CU / Cloud (chipsets / GPUs, accelerators, CaaS) .............. 31
6.2.1 Vodafone perspective.......................................................... 31
6.2.2 O-RAN specifications ........................................................... 31
6.2.3 Different options ................................................................. 32
6.2.4 Collaboration test setup ...................................................... 32
6.2.5 Gains and achievements in energy savings optimization
research ........................................................................................... 33
6.2.6 Measurement techniques ................................................... 35
6.2.7 Findings and results ............................................................. 38
6.2.8 Future work and next steps ................................................. 40
6.2.9 Recommendations ............................................................... 40
6.3 RIC applications (xApps/rApps), SMO ............................................ 41
6.3.1 Operator perspective........................................................... 41
6.3.2 O-RAN specifications: architecture and procedure flow for
cell on / off use case ........................................................................ 41
6.3.3 Two options for notification from ES-rApp to TS-xApp ....... 42
6.3.4 Collaboration test setup ...................................................... 43

5
Version 1.0
Open RAN Handbook 2nd Edition
6.3.5 Keysight RICtest configuration and statistics ...................... 44
6.3.6 Procedural flows .................................................................. 45
6.3.7 Findings and results ............................................................. 48
6.3.8 Recommendations ............................................................... 50

7 Summary and Recommendations ................................................. 52

8 References................................................................................... 54

9 Glossary of Terms ........................................................................ 55

6
Version 1.0
Open RAN Handbook 2nd Edition
List of Figures

Figure 1. 3GPP release 17 - 19 energy-saving features ............................................................................................... 11

Figure 2. O-RAN ALLIANCE network energy savings feature planning overview ......................................................... 14

Figure 3. O-RAN architecture (image provided by the O-RAN ALLIANCE) .................................................................. 15

Figure 4. O-RAN components ...................................................................................................................................... 16

Figure 5. Interfaces available for the base station....................................................................................................... 21

Figure 6. Available interfaces for the O-RAN radio unit .............................................................................................. 21

Figure 7. Low load resource allocation ........................................................................................................................ 22

Figure 8. Medium load resource allocation ................................................................................................................. 22

Figure 9. Busy load resource allocation ....................................................................................................................... 22

Figure 10. Full load resource allocation ....................................................................................................................... 22

Figure 11. Keysight O-RU automated energy test and reporting solution .................................................................. 25

Figure 12. Keysight O-RU automated energy test flow ............................................................................................... 26

Figure 13. C-States (image provided by Intel).............................................................................................................. 32

Figure 14. Topology of the collaboration test setup for O-RAN PlugFest Spring 2022 ................................................ 34

Figure 15. Keysight energy plane test suite topology .................................................................................................. 35

Figure 16. UE groups with different path loss and traffic load .................................................................................... 36

Figure 17. Automated test iterations........................................................................................................................... 36

Figure 18. Test report and chart .................................................................................................................................. 38

Figure 19. Multi-vendor rApp-xApp coordination topology (image provided by Juniper Networks) .......................... 43

Figure 20. Keysight RICtest map designer.................................................................................................................... 44

Figure 21. Keysight RICtest cell configuration ............................................................................................................. 45

Figure 22. Keysight RICtest statistics dashboard ......................................................................................................... 45

Figure 23. Option 1 topology (image provided by Rimedo Labs) ................................................................................ 46

Figure 24. Option 2 topology (image provided by Rimedo Labs) ................................................................................ 47

Figure 25. Number of cells across different frequency bands and sites in the Vodafone dataset .............................. 49

Figure 26. Traffic load, prediction, and energy-saving modes over time with AirHop’s ESMC rApp (image provided
by AirHop Communications) ........................................................................................................................................ 49

7
Version 1.0
Open RAN Handbook 2nd Edition

List of Tables

Table 1. Energy savings measurements ....................................................................................................................... 17

Table 2. Band 1 – static power consumption and RF output power............................................................................ 27

Table 3. Band 1 – average power consumption reporting .......................................................................................... 28

Table 4. Band 1 – daily energy consumption reporting ............................................................................................... 28

Table 5. Band 3 – static power consumption and RF output power............................................................................ 28

Table 6. Band 3 – average power consumption reporting .......................................................................................... 29

Table 7. Band 3 – daily energy consumption reporting ............................................................................................... 29

Table 8. Band 75 – static power consumption and RF output power ......................................................................... 29

Table 9. Band 75 – average power consumption reporting ........................................................................................ 30

Table 10. Band 75 – daily energy consumption reporting ........................................................................................... 30

Table 11. Energy Savings Optimization Table .............................................................................................................. 33

Table 12. components collaborators at the O-RAN PlugFests Spring / Fall 2022 and at MWC 2023 .......................... 34

Table 13. Sample of the Test Report ........................................................................................................................... 37

Table 14. Test report with energy-saving gain ............................................................................................................ 38

8
Version 1.0
Open RAN Handbook 2nd Edition

1 Executive Summary
Network energy efficiency is a key sustainability goal for mobile network operators (MNOs) around the
world aiming to reduce environmental impacts and achieve operational cost savings, as energy
consumption is a major component of their operational expense (OpEx). European energy costs have
risen by more than 400 percent, on average, over a two-year period. While many MNOs’ individual OpEx
cost components have dropped over a similar period, energy costs have risen by 2 percent of overall
OpEx cost and are expected to continue to increase.

Against this context, Vodafone has set ambitious targets to reduce its carbon footprint. It aims for net-
zero emissions across its entire value chain by 2040 and, in the near term, powering all operations with
renewable energy by 2025.

The radio access network (RAN) consumes most of the network energy (~70%) and is therefore the focus
of optimization for monolithic base stations in terms of energy consumption, efficiency, and artificial
intelligence (AI) / machine learning-enabled (ML) savings techniques.

While the Open Radio Access Network (Open RAN) is an attractive option for operators with key benefits
such as supplier diversification and liberating AI / ML innovation, it remains necessary to demonstrate
that its energy efficiency and savings aspects are at least on par with or better than those of monolithic
RANs.

Vodafone has been a leader in advancing Open RAN deployments across its markets, with live sites
currently operating in countries, including the UK and Romania, with the potential to deploy in other
partner markets. These deployments demonstrate Vodafone's commitment to diversifying the vendor
ecosystem and achieving greater network flexibility and efficiency. In addition, the key performance
indicators (KPIs) of these sites now demonstrate that the performance of Open RAN exceeds that of the
legacy equipment in most areas, including 4G and 5G call success rates, as well as download and upload
speeds across multiple spectrum frequencies.

Since 2022, Vodafone has taken a pivotal step to evaluate and test the energy efficiency of Open RAN
solutions by hosting a series of focused PlugFests. These events gathered multiple vendors and fostered
collaboration to identify and develop innovative energy optimization techniques for Open RAN
deployments.

In this handbook, Vodafone and Keysight will share:

• Progress on the state of the 3GPP and O-RAN ALLIANCE standards with respect to Open RAN energy
savings features development.

• Open RAN energy consumption, efficiency, and savings benchmarking considerations, test
scenarios, methodologies, and test setups for the Open Radio Unit (O-RU), Open Distributed Unit
/ Central Unit / Cloud (O-DU / O-CU / O-Cloud), RAN Intelligent Controller (RIC), and the associated
xApps / rApps.

• Test results and key observations for the testing conducted with Vodafone’s technology partners
in the Vodafone labs, which started in 2022 up until the end of 2024.

• Recommendations for improving standards and product implementations.

9
Version 1.0
Open RAN Handbook 2nd Edition

2 Motivation and Challenges

Energy efficiency is a critical consideration in the evolution of Open RAN technology due to its significant
impact on sustainability, operational costs, and overall network performance. The drive for greater
energy efficiency stems from the increasing demand for data services, the deployment of higher-capacity
networks like 5G, and the pressing need for environmentally sustainable solutions. Open RAN introduces
a modular and disaggregated architecture that fosters supplier diversity and innovation, enabling
collaboration across a diverse ecosystem of hardware and software providers. This diversity creates an
exciting opportunity to pioneer advanced energy-saving techniques and establish benchmarks that can
surpass traditional monolithic RAN systems.

Testing and validating energy efficiency in Open RAN components such as the O-RU, O-DU, O-CU, and
the RIC involve addressing varying operational scenarios, traffic patterns, and deployment configurations
that reflect real-world conditions. While promising in terms of enhancing efficiency, the inclusion of AI /
ML-driven features adds layers of complexity to validation, requiring robust frameworks to assess the
real-world effectiveness of such features. Furthermore, the dynamic nature of Open RAN networks,
where components must adapt to changing workloads and environmental conditions, requires advanced
testing methodologies that go beyond traditional lab environments to include real-world operational
contexts. These advancements align with the industry’s broader shift toward automation and intelligent
systems, creating a robust foundation for energy-efficient, future-ready networks.

Another challenge is aligning vendor-specific implementations with global standards. While organizations
such as the 3GPP and O-RAN ALLIANCE are actively developing energy-saving features, the varying pace
of adoption and differences in implementation across vendors create discrepancies in energy
performance. These discrepancies highlight the need for clear, standardized benchmarks and
collaborative efforts among stakeholders to harmonize approaches.

The collaborative efforts of global standards bodies like 3GPP and the O-RAN ALLIANCE should further
increase the potential for unified approaches to energy efficiency. While vendor-specific
implementations vary, the diversity presents an opportunity to drive alignment and innovation through
collective efforts, ensuring Open RAN solutions deliver consistent and measurable energy savings.

The continued focus on energy efficiency testing, benchmarking, and real-world validation is not only a
technical imperative but also an opportunity to set new industry standards. By embracing these
opportunities, Open RAN can confidently demonstrate its ability to meet and exceed operators' energy
efficiency expectations, paving the way for a more sustainable and economically efficient future for
mobile networks.

10
Version 1.0
Open RAN Handbook 2nd Edition

3 Standardization and Industry Forums

The 3GPP and O-RAN ALLIANCE are actively developing specifications for radio access network energy
savings.

The 3GPP focuses on the air interface and NG-RAN aspects of the network energy savings studies and
normative specifications. Refer to Figure 1.

The O-RAN ALLIANCE focuses on specifying the requirements, architecture, procedures, protocols, and
test specifications required to support interoperable, multi-vendor Open RAN functions for energy
savings features and capabilities.

3.1 3GPP

Figure 1. 3GPP release 17 - 19 energy-saving features

In Release 17, the 3GPP initiated a study item (3GPP RP-201304) on AI / ML for NG-RAN with network
energy savings, one of the three use cases described in 3GPP TR 37.817 [1] which summarizes the output
of this study.

Cell activation / deactivation is an energy-saving scheme in the spatial domain that exploits traffic
offloading in a layered structure to reduce the energy consumption of the entire RAN. When the expected
traffic volume is lower than a fixed threshold, the cells may be switched off, and the served user
equipment (UEs) may be offloaded to a new target cell.

It is also possible to achieve efficient energy consumption by other means, such as reducing load,
coverage modification, or making other RAN configuration adjustments. The optimal energy-saving
decision depends on many factors, including the load situation at different RAN nodes, RAN nodes'
capabilities, KPI / Quality of Service (QoS) requirements, number of active UEs and UE mobility, cell
utilization, and more.

11
Version 1.0
Open RAN Handbook 2nd Edition
However, identifying actions aimed at energy efficiency improvements is not trivial. The wrong switch-
off of the cells may seriously deteriorate the network performance since the remaining active cells need
to serve the additional traffic. Incorrect traffic offload actions may lead to a deterioration of energy
efficiency instead of an improvement.

The conventional network energy-saving techniques are vulnerable to potential issues such as:

1. Inaccurate cell load prediction: Energy-saving decisions currently rely on the current traffic load
without considering future traffic load.

2. Conflicting targets between system performance and energy efficiency: Maximizing the system’s KPIs
is usually done at the expense of energy efficiency. Similarly, the most energy-efficient solution may
impact system performance. Thus, there is a need to balance and manage the trade-off between the
two.

3. Conventional energy-saving-related parameters adjustment: Energy-saving-related parameters

configuration is set by traditional operation. This is based on different thresholds of cell load for cell
switch on / off, which is somewhat rigid since it is difficult to set a reasonable threshold.

4. Conflicting goals in optimizing both local and overall energy efficiency: Actions that may produce a
local (for example, limited to a single RAN node) improvement in energy efficiency, could
simultaneously result in an overall (for example, involving multiple RAN nodes) deterioration of
energy efficiency. As a result, it is important to find a balance between the two.

To confront the issues listed above, it is possible to use AI / ML techniques to optimize energy-saving
decisions by leveraging the data from the RAN network. AI / ML algorithms may predict the energy
efficiency and load state of the next period, providing useful information to make better decisions on cell
activation / deactivation for energy savings. Based on the predicted load, the system may dynamically
configure the energy-saving strategy; for example, the switch-off timing, granularity, and offloading
actions, to maintain a balance between system performance and energy efficiency and reduce energy
consumption.

The normative work based on the conclusion of this Release 17 study item was undertaken in Release 18
(3GPP RP-233441), with specifications of data collection enhancements and signaling support within
existing NG-RAN interfaces and architecture (including non-split architecture and split architecture) for
AI / ML-based network energy savings, load balancing, and mobility optimization.

The 3GPP has approved a study item in Release 19 to study enhancements for AI / ML for NG-RAN, which
include energy-saving enhancements, such as energy cost prediction.

In Release 18, the 3GPP completed a study (3GPP RP-220297) on network energy savings for NR. It has
studied, identified, and evaluated techniques on the gNB and UEs to improve energy savings of the
networks for both FDD and TDD, FR1 and FR2. The output of this study is summarized in 3GPP TR 38.864
[2]. These techniques were evaluated based on the base station (BS) energy consumption model,
evaluation methodology, and KPIs from multiple domains which were used to assess and optimize energy
consumption and savings gain. At the same time, these techniques aim to find the optimal balance on
the impacts to network and user performances (such as spectral efficiency, capacity, UPT, latency,
handover performance, call drop rate, initial access performance, SLA assurance related KPIs), energy
efficiency, UE power consumption, and complexity.

The study prioritizes idle / empty and low / medium load scenarios, allows different loads among carriers
and neighbor cells, and permits legacy UEs to continue accessing a network implementing Rel-18 network
energy savings techniques, except for techniques developed specifically for greenfield deployments.

12
Version 1.0
Open RAN Handbook 2nd Edition
The evaluation of energy savings techniques explored time, frequency, spatial, and power optimizations.

• The techniques in time and frequency domains mainly aim to reduce the power consumption for
the dynamic power part, by trying to shut down more symbols on one or more carriers to achieve
BS micro sleep, and even for the static power part, by enlarging the interval between the
contiguous active transmission / reception occasions to achieve BS light / deep sleep.

• The techniques in spatial and power domains mainly aim to reduce the power consumption of the
transceiver (TRX) chains and power amplifiers (PAs) by trying to shut down more spatial elements
and / or reduce transmission power / power spectrum density or increase PA efficiency.

Based on the study’s findings, the 3GPP approved a work item on network energy savings for NR in
Release 18 (3GPP RP-230566) and has developed normative specifications in Release 18 for a set of
network energy savings techniques. The specified techniques include:

• serving-cell without synchronization signal block (SSB-less SCell) operation

• enhancement of the cell DTX / DRX mechanism

• methods for the spatial and power domains to enable efficient adaptation of:

o spatial elements (such as antenna ports, active transceiver chains)

o power offset values between physical downlink shared channel (PDSCH) and channel state
information reference signal (CSI-RS)

• cell handovers procedure enhancement(s)

• inter-node beam activation and enhancements among others.

In Release 19, the 3GPP has approved a work item for enhancements of network energy savings for NR
(3GPP RP-240170) to specify further network energy savings targeting the beneficial techniques studied
in Rel-18, but unspecified. These techniques include on-demand SSB and on-demand system information
block type 1 (SIB1) transmissions, as well as adaptation of common signal / channel transmissions

3.2 O-RAN ALLIANCE

The O-RAN ALLIANCE recently completed its specifications for Release 004 features (see Figure 2),
including RIC AI / ML-enabled network energy savings features and capabilities. The features listed below
are fully compatible with 3GPP specifications:

• The RF cell/carrier on / off complements the 3GPP Release 18-specified NG-RAN feature for cell /
carrier on / off.

• The RF channel reconfiguration enables selective on / off switching of antenna elements on the
massive multiple-input / multiple-output (MIMO) antenna array.

• The RF advanced sleep mode enables on / off switching for slots and symbols.

Release 005 candidates (see Figure 2) include:

13
Version 1.0
Open RAN Handbook 2nd Edition
• O-Cloud energy savings, which involve switching on / off computing resources (cores) and adjusting
their sleep modes / levels, such as a CPU’s C-States / P-States.

• Alignment with the 3GPP Release 18-specified network energy savings features

• Test specifications.

Figure 2. O-RAN ALLIANCE network energy savings feature planning overview

14
Version 1.0
Open RAN Handbook 2nd Edition

4 Open RAN Architecture and Components

The O-RAN ALLIANCE specifies the Open RAN architecture [3] (refer to Figure 3), as consisting of
components that implement O-RAN standards-compliant interfaces.

Figure 3. O-RAN architecture (source O-RAN ALLIANCE)

O-RAN-compliant components (refer to Figure 4) include:

• O-RU: Different categories of radio A, B, and ULPI Class A / B.

• O-DU / CU: RAN workloads that operate on the Cloud infrastructure.

• RIC and RIC applications: AI-enabled xApps / rApps, and service management and orchestration
(SMO).

• O-Cloud: Chipsets / GPUs, accelerators, Cloud as a Service (CaaS) sub-components, and server
infrastructure.

15
Version 1.0
Open RAN Handbook 2nd Edition

Figure 4. O-RAN components

Various suppliers can implement each of these O-RAN components using different options. For example,
choose from several options to implement the O-Cloud. This requires evaluating performance gains
versus CapEx, OpEx, and complexity without compromising customers’ experiences and service level
agreements (SLAs).

Here is an example of a listing of computing platform options and considerations:

• Different processing architectures (x86 and/or GPUs)

• Different acceleration options (inline versus look aside)

• Different chipsets (Intel x86, NVIDIA, AMD, ARM / Marvell, and more)

• Different vendors and virtualization stacks versus bare metal

• Different vDU / vCU energy savings interoperability with O-Cloud

• Different disaggregation strategies.

• Use the same vendor with different configurations, software / firmware / hardware.

Each component and the entire Open RAN solution as a system under test must be performance-tested
and benchmarked using a well-defined, consistent, and repeatable testing methodology. This is to ensure
proper evaluation and optimization of the Open RAN solution for multi-vendor interoperability with high
performance, which is either on par with or better than traditional RAN solutions in commercial networks
today.

16
Version 1.0
Open RAN Handbook 2nd Edition

5 Open RAN Energy / Performance Benchmarking

Considerations
Open RAN energy tests and benchmarking must consider energy consumption, energy efficiency, and
energy savings.

• Energy consumption (EC) is the measured energy used by the DUT / SUT; unit is in Joules (J) or
Kilowatt-hour (kWh).

• Energy efficiency (EE) is the measured effectiveness of energy consumed based on certain factors
such as data volume (DV), coverage, and network slices. For example, compute EE (DV) for bi-
directional data volume delivered per unit of energy consumed (bits/J).

• Energy savings (ES) refers to dynamic closed-loop optimization of energy consumption and
efficiency leveraging AI / ML-assisted methods and AI / ML models. The AI / ML models require
training and testing and must be monitored throughout their life cycle.

It is possible to measure and evaluate energy savings using a set of multidimensional KPIs that seek to
maximize energy savings (or minimize losses). This is accomplished while considering material costs and
complexity, and ensuring no compromise (or acceptable planned impacts) to the user experience (user
QoS), service level agreements, and network performances (network KPIs).

Table 1 is a sample list of measurements applicable for energy savings performance evaluation.

Table 1. Energy savings measurements

Measurements Measurements Comments

Category

1) Energy a) Energy consumption Network energy saving gain is

b) Energy efficiency (data volume / computed based on the energy
energy consumed) consumption for a technique and
c) Energy savings gain / loss the baseline over the same
duration.
2) User a) User perceived throughput (UPT)
experiences gain / loss
b) Packet losses, jitter, and latency
3) Network a) Initial access performance (success
performance rate, and latency)
b) Call drop rate
c) Cell block error rate (BLER)
d) Cell throughput
e) Cell retransmissions
f) Handover performance (success
rate)
g) Spectral efficiency

17
Version 1.0
Open RAN Handbook 2nd Edition

4) Computing a) CPU utilization (number of cores Computing efficiency is less

and % utilization) straightforward in relation to
b) Memory utilization (%) differences in computing
platforms (differences in
accelerators, chipsets, and more).
Material costs may also require
consideration.

18
Version 1.0
Open RAN Handbook 2nd Edition

6 Open RAN Energy Benchmarking Test Scenarios,

Methodology and Setups
Vodafone has set challenging targets for energy-saving goals and reducing its carbon footprint. The
company aims to achieve net-zero emissions across its entire value chain by 2040, with a near-term goal
of cutting its environmental impact in half by 2030. In line with this vision, energy efficiency in Open RAN
networks is critical to ensuring a sustainable and scalable telecommunications infrastructure.

Open RAN offers a unique opportunity to integrate innovative energy-saving mechanisms across
disaggregated components, such as O-RUs and cloud-based O-DU / CU systems, through rigorous testing
and optimization. Vodafone is exploring groundbreaking methods to automate energy efficiency,
dynamically adjusting network parameters based on traffic patterns and environmental conditions by
leveraging advanced xApps and rApps deployed via the RIC platform. These efforts align with Vodafone’s
sustainability goals and also set a benchmark for the telecommunications industry by demonstrating how
technology can address the twin imperatives of operational excellence and environmental responsibility.

In the following sections, we will delve into the specifics of various testing activities conducted in
Vodafone’s labs, showcasing our innovative approach to enhancing energy efficiency in Open RAN
networks. Additionally, we will highlight a key initiative to which Vodafone actively contributed by
providing its dataset to demonstrate tangible energy savings, further solidifying its commitment to
driving sustainability in telecommunications.

6.1 O-RU
6.1.1 Vodafone perspective
As the most power-hungry component in the RAN, O-RUs are central to energy transformation due to
the high demands of the PA. The energy O-RUs consume not only exceeds that of other network elements
but also directly impacts both operational costs and environmental sustainability. As the O-RAN ALLIANCE
standardizes energy-saving features and vendors work to implement these innovations, assessing the
energy efficiency of O-RUs is now paramount for MNOs.

When evaluating O-RU energy performance, Vodafone focuses on two key areas:

• First, we assess pure hardware consumption — measuring the O-RU’s power draw without any
active energy-saving features to establish a baseline.

• Second, we evaluate performance with energy-saving features (ESFs) activated, such as PA

shutdown during idle PDSCH symbols, which can significantly lower power demands without
compromising service quality.

Rigorous benchmarking in these areas enables Vodafone to compare vendor solutions, ensuring
alignment with energy efficiency standards while optimizing network sustainability. Through these
efforts, Vodafone is a leader in sustainable connectivity, exemplifying responsible innovation and building
a greener, more efficient future for telecommunications.

19
Version 1.0
Open RAN Handbook 2nd Edition

6.1.2 O-RAN specifications

The O-RAN ALLIANCE established several working groups (WGs) to define the various interfaces between
diverse elements in the disaggregated radio access network. One of them, WG4, focuses on the definition
of the fronthaul interface between O-RU and O-DU [4].

• The O-RU includes most of the low-level 3GPP physical layer (L1) operations, such as FFT / IFFT,
filtering, up / down conversion, beamforming, and other radio frequency processes. It is the
element that interfaces with mobile users over the air.

• The O-DU includes most of the high-level physical layer or most of the baseband processing of the
communications.

The O-RAN WG4 conformance test specification [5] defines the testing required for certification of the
O-RU and O-DU as WG4 fronthaul-compliant. The test specification includes different chapters:

1. Device under test classification:

a. Standalone testing for O-RU

b. Standalone testing for O-DU

2. WG4 fronthaul planes:

a. M-Plane tests
b. CU-Plane tests
c. S-Plane tests

To validate the energy-saving capabilities declared by an O-RU, the O-RAN WG4 conformance test
specification provides several test cases in the M-Plane and CU-Plane chapters dedicated to energy-
saving features that the M-Plane and CU-Plane control. The main objective of these tests is to validate
that the O-RU applies the energy-saving features that it declares rather than its efficiency in energy-
saving. This means the WG4 O-RAN conformance test specification does not target the energy
consumption characterization of the O-RU but rather functional testing.

Also the ETSI 202 706 [6] test specification attempts to test for actual energy consumption and
characterization. This test specification originally defines a set of tests that aims to characterize the
power consumption of a BS under static conditions (refer to Figure 5), rather than an O-RU in isolation.
Although this test specification targets the BS, it allows for distributed BS models. In a distributed BS the
main metrics are measured at the RF interface (antennas), and at the direct current (DC) or alternate
current (AC) feed entry of the distributed BS (power feed). The O-RU (refer to Figure 6) includes both
interfaces so the ETSI test specification can be reused to test the O-RU.

20
Version 1.0
Open RAN Handbook 2nd Edition

Figure 5. Interfaces available for the base station

Figure 6. Available interfaces for the O-RAN radio unit

The ETSI test specification describes several loads that define the amount of resource allocation in the
grid to which the BS, or the O-RU, schedules and transmits during the test. Therefore, these are purely
downlink tests. These loads are:

21
Version 1.0
Open RAN Handbook 2nd Edition
1. Low load: This load uses a small set of resource allocations for SSB and remaining minimum system
information (RMSI) (refer to Figure 7).

Figure 7. Low load resource allocation

2. Medium load: This load uses an increased set of resource allocations for SSB and 30 percent of the
allocation available for PDSCH in the radio frame (refer to Figure 8).

Figure 8. Medium load resource allocation

3. Busy load: This load also uses SSB allocation, and 50 percent of the allocation is available for PDSCH
in the radio frame (refer to Figure 9).

Figure 9. Busy load resource allocation

4. Full Load: This load uses the full grid such as the NR-TM1.1 test vector. There is no SSB allocation
for this load (refer to Figure 10).

Figure 10. Full load resource allocation

22
Version 1.0
Open RAN Handbook 2nd Edition

The goal is to identify the energy consumption of the O-RU under these different loads for comparison
purposes. This test assumes constant data transmission over the load interval. The test durations defined
in the specifications target two types of power consumption measurements:

1. Average power consumption: The target characterizes the average power consumed by the BS, or
O-RU in this case, over an equal period for each load.
2. Daily consumption: The target characterizes the BS's, or O-RU's daily power consumption over a 24-
hour period. During this 24-hour period, three different load conditions — low load, medium load,
and busy load — are tested with different time durations to mimic different traffic demands during
the day.

The test results are published in test reports, which provide network vendors and operators with key
metrics such as RF power output and DC / AC power consumption under different conditions.

6.1.3 Different options

There are several options for implementing energy-saving features for O-RUs. Some are proprietary to
the vendor, meaning that the algorithms are vendor-dependent and not defined in any standard, and
some are standard as defined by the O-RAN WG4 fronthaul specification:

1. Proprietary algorithms are a suitable path for the early implementation of energy-saving features in
radios. This is because the standard might take some time to reach the desired level of maturity.
Proprietary implementations might also help shape the specification's future and provide excellent
insight into what works best.

As the vendor implementation does not necessarily follow a standard or an industrial agreement, the
algorithms used are based on autonomous operation. This means the O-RU vendor must implement
energy-saving mechanisms that operate without requiring any signalling or external input — for
example, from an O-DU or other device in the network. Some examples of autonomous vendor
implementations are the temporary switching on and off of unused antennas and RF front elements
such as power amplifiers (PAs) when there is no temporary allocation of traffic that requires these
resources. In addition, radios might independently decide to go into deep sleep for an undetermined
amount of time if no traffic is detected and wake up when necessary.

2. The O-RAN WG4 fronthaul specification defines non-real-time operations (M-Plane) [7] and real-time
operations (CU-Plane) [1]. Both allocate capabilities to manage energy consumption and savings in
the O-RU.

M-Plane energy-saving operations target macro changes in the O-RU to substantially decrease energy
consumption. In this case, the O-DU actively sets some radio settings in energy-saving mode.
Examples include switching off entire antenna panels, some antenna elements, carriers, bands, and
sectors. Another example is when the O-DU might decide to put parts of or the entire O-RU to sleep
if no traffic is expected to be sent to the radio. These operations take a long time to execute and
more importantly, take a long time to turn back on the parts of the O-RU that were set off or that
went to sleep.

23
Version 1.0
Open RAN Handbook 2nd Edition
CU-Plane energy-saving operations target micro changes in the O-RU operation and typically for a
shorter time period. Like M-Plane-based energy savings, the O-DU actively signals the radio what
actions the radio needs to perform with the intention of lowering energy consumption. The only
difference here is timing restrictions. The changes are signaled in real time and can be applied as
briefly as an orthogonal frequency division multiplexing (OFDM) single symbol. For example, the O-
RU might dynamically switch off power amplifiers or antenna elements if it detects no fronthaul
traffic is scheduled for the radio in those antennas. Likewise, the O-DU might decide that fewer
antenna elements are needed to serve users but consider that the situation might change and does
not commit to putting parts or the entire O-RU to sleep.

As an initial attempt to cover energy-saving features for O-RU, the C-Plane section type 0 (ST0)
messaging was defined in the O-RAN WG4 control plane, user plane, and synchronization plane (CUS-
Plane) specification to enable an O-DU to explicitly signal to an O-RU that certain resource allocations
for a determined period must not be allocated, or in other words, silenced. This could be used to
allow an O-RU to run internal calibrations or to execute energy consumption algorithms, such as
switching off radio elements. Although this ST0 could be useful for energy saving, it is not very flexible
and is limited in scope. For this reason, newer versions of the specification have added new section
types to expand this. Section Type 4 and Section Type 8 messages were introduced as a new method
for the O-DU to explicitly signal to the O-RU which energy-saving methods the radio must apply. For
example, Section Type 4 messages with the transceiver control (TRX_Control) command type allow
the O-DU to enable and disable a specific number of antenna elements when using a set of resource
allocations with a determined or undetermined duration. Another example is Section Type 4
messages with advanced sleep mode; the ASM command type allows the O-DU to place the O-RU in
certain degrees of sleep according to a specific or unspecific amount of time, as needed.

Although the goal tends to standardize as many features as possible, it also leaves room for network
equipment vendors to innovate and implement autonomous, proprietary implementations. These
incentives push radio vendors to bring differential features into a competitive environment and are
especially appealing to network operators. At the same time, the standardization bodies continue to
define requirements for a minimal set of operations.

6.1.4 Collaboration test setup

The device under test, the O-RU, provides three important probe locations:

1. RF antenna ports: The deployment of the radio units includes antenna panels. To test them, there
are two options:
a. OTA: The radio signals travel from and into the radio; the testing requires environmentally
and electromagnetically-controlled chambers and test probe antennas.
b. Conducted: The radio signals travel over a conductor, such as cabling. In this case, testing is
simplified, as only cables connecting the radio to the test equipment are necessary.
2. Optical fronthaul interface: This interface connects the radio to one or multiple O-DUs, either as a
point-to-point connection or via a network. Typically, this connection is made via optical cabling. For
testing, the O-RU connects to an O-DU emulator that provides O-RAN traffic.
3. DC power feed: Voltage and current feed into the radio's power supply unit to power it up. The testing
equipment provides the demanded voltage and current and can report these values.

24
Version 1.0
Open RAN Handbook 2nd Edition

All three of these interfaces connect to the test equipment, and measurements can be taken to
characterize the power transmitted by the O-RU and to monitor the DC power consumption
simultaneously.

In the tests run between Vodafone and Keysight, the solution used is a combination of radio equipment,
network emulation, and power and energy test equipment, all integrated under Keysight’s unique test
automation umbrella (refer to Figure 11).

Figure 11. Keysight O-RU automated energy test and reporting solution

The test solution consists of the following components:

• Power sensor: Measures the RF power detected at the output of the radio.

• Switch matrix: Enables the connection of multiple antenna ports to single or multiple power
sensors.

• O-DU emulator: Generates and captures O-RAN traffic to and from the radio over the fronthaul.

• DC power supply: Feeds the voltage and current into the radio to power it up and reports
consumption of the voltage and current.

• Test automation execution software: Coordinates the full execution of the energy consumption
tests by integrating the control of all the hardware and software involved in the solution.

• Test report and analytics: Provides the final report and insights on the measurements taken during
the test.

25
Version 1.0
Open RAN Handbook 2nd Edition

Figure 12. Keysight O-RU automated energy test flow

Here are the steps in the test flow that appear in Figure 12:

1. Initialize the test case: Run pretest operations such as establishing connections and selecting test
parameters.
2. Set up test equipment: Configure the test equipment including the power supply, power sensors, and
O-DU emulator.
3. Test each radio band: If the radio supports multiband, iterate the test for each band, one at a time.
4. Test each load: The test covers the different load configurations needed for each band. Depending
on the test, not all the loads are used.
5. Test each frequency: For each band and load, the test goes over the different frequencies within the
band such as the low, middle and high frequency parts within each band. Depending on the load, not
all the frequencies are tested. For example, low load does test all three frequencies. For all other
loads, only the middle frequency in the band is tested.
6. Generate test vectors: This occurs according to the load distribution under test.
7. Send O-RAN traffic: The O-RU processes the O-RAN traffic and transmits the RF signal accordingly.
8. Make measurements: Measure both DC power and RF output power while the radio transmits the
signals. If there are multiple antennas, measure the RF power output for each antenna.
9. Data collection and storing: The test gathers data for each iteration.
10. Test report: After completing all the iterations, the test case generates a test report per the ETSI
specification.

26
Version 1.0
Open RAN Handbook 2nd Edition

6.1.5 Findings and results

The main tests executed on the Fujitsu O-RU1 are the average power consumption (APC) and the daily
energy consumption (DEC) described in [6]. Both tests comply with the following conditions and
requirements:

1. For APC the measurement unit is in watts (W).

2. For DEC, the measurement unit is measured in watts per hour (Wh).
3. The following loads run in these two tests:
a. Low load with a 2-hour duration for each frequency within a band, for a total duration of 6
hours.
b. Medium load with only a single frequency, the middle frequency, for a duration of 10 hours.
c. Busy hour load with only a single frequency, the middle frequency, for a duration of 8 hours.
d. Full load with only a single frequency, the middle frequency, for a duration of 2 hours.
4. For each antenna port, the measurement unit for radio frequency output power is in watts (W).
a. This measurement is run for all four loads — low, medium, busy hour, and full load.
b. Full load is set to 2 hours of test duration.
c. The output power that appears in the test report is the average RF output across all antenna
ports.
5. If an O-RU supports multiple bands, the test is repeated for each band, one at a time. A test report is
generated individually for each band.

The Fujitsu O-RU provides the following test reports for the three bands it supports (refer to Table 2 to
10).

Band 1 - NR:

Table 2. Band 1 – static power consumption and RF output power

4 Static power consumption (measured)

4.1 Full load, middle-frequency channel 453.51 W
4.2 Busy hour load, middle-frequency channel 335.8 W
4.3 Medium load, middle-frequency channel 277.86 W
4.4 Low load
4.4
.1 Low-end frequency channel 209.61 W
4.4
.2 Middle-frequency channel 209.64 W
4.4
.3 High-end frequency channel 209.23 W
4.4
.4 Average consumption with low load 209.49 W
5 RF output power
5.1 Full load, middle-frequency channel 16.99 W
5.2 Busy hour load, middle-frequency channel 9.13 W

1 A model provided for proof-of-concept testing in Vodafone’s Open RAN central lab

27
Version 1.0
Open RAN Handbook 2nd Edition

5.3 Medium load, middle-frequency channel 5.39 W

5.4 Low load
5.4
.1 RF output power at low-end channel 0.76 W
5.4
.2 RF output power at middle-end channel 0.73 W
5.4
.3 RF output power at high-end channel 0.77 W
5.4
.4 Average RF output power with low load 0.75 W

Table 3. Band 1 – average power consumption reporting

Average power consumption reporting

S.No Parameter Value Unit
1 Pequipment of integrated RU power consumption W
3 Pequipment of distributed RU power consumption
Pequipment of distributed RU power consumption for the central
3.1 part W
3.2 Pequipment of distributed RU power consumption for remote parts 289.58 W

Table 4. Band 1 – daily energy consumption reporting

Daily energy consumption reporting

S.No Parameter Value Unit
1 Eequipment of integrated RU energy consumption Wh
3 Eequipment of distributed RU energy consumption
Eequipment of distributed RU energy consumption for the central
3.1 part Wh
3.2 Eequipment of distributed RU energy consumption for remote parts 5212.24 Wh

Band 3 - NR:

Table 5. Band 3 – static power consumption and RF output power

4 Static power consumption (measured)

4.1 Full load, middle-frequency channel 531.43 W
4.2 Busy hour load, middle-frequency channel 383.88 W
4.3 Medium load, middle-frequency channel 305.53 W
4.4 Low load
4.4
.1 Low-end frequency channel 213.61 W
4.4
.2 Middle-frequency channel 213.6 W

28
Version 1.0
Open RAN Handbook 2nd Edition

4.4
.3 High-end frequency channel 213.55 W
4.4
.4 Average consumption with low load 213.59 W
5 RF output power
5.1 Full load, middle-frequency channel 26.82 W
5.2 Busy hour load, middle-frequency channel 15.17 W
5.3 Medium load, middle-frequency channel 8.94 W
5.4 Low load
5.4
.1 RF output power at low-end channel 1.22 W
5.4
.2 RF output power at the middle-end channel 1.22 W
5.4
.3 RF output power at high-end channel 1.22 W
5.4
.4 Average RF output power with low load 1.22 W

Table 6. Band 3 – average power consumption reporting

Average power consumption reporting

S.No Parameter Value Unit
1 Pequipment of integrated RU power consumption W
3 Pequipment of distributed RU power consumption
Pequipment of distributed RU power consumption for the central
3.1 part W
3.2 Pequipment of distributed RU power consumption for remote parts 321.43 W

Table 7. Band 3 – daily energy consumption reporting

Daily energy consumption reporting

S.No Parameter Value Unit
1 Eequipment of integrated RU energy consumption Wh
3 Eequipment of distributed RU energy consumption
Eequipment of distributed RU energy consumption for the central
3.1 part Wh
3.2 Eequipment of distributed RU energy consumption for remote parts 5785.69 Wh

Band 75 - NR:

Table 8. Band 75 – static power consumption and RF output power

4 Static power consumption (measured)

4.1 Full load, middle-frequency channel 553.98 W

29
Version 1.0
Open RAN Handbook 2nd Edition

4.2 Busy hour load, middle-frequency channel 396.65 W

4.3 Medium load, middle-frequency channel 312.51 W
4.4 Low load
4.4
.1 Low-end frequency channel 211.83 W
4.4
.2 Middle-frequency channel 211.68 W
4.4
.3 High-end frequency channel 211.6 W
4.4
.4 Average consumption with low load 211.7 W
5 RF output power
5.1 Full load, middle-frequency channel 25.28 W
5.2 Busy hour load, middle-frequency channel 14.16 W
5.3 Medium load, middle-frequency channel 8.35 W
5.4 Low load
5.4
.1 RF output power at low-end channel 1.12 W
5.4
.2 RF output power at the middle-end channel 1.12 W
5.4
.3 RF output power at high-end channel 1.13 W
5.4
.4 Average RF output power with low load 1.12 W

Table 9. Band 75 – average power consumption reporting

Average power consumption reporting

S.No Parameter Value Unit
1 Pequipment of integrated RU power consumption W
3 Pequipment of distributed RU power consumption
Pequipment of distributed RU power consumption for the central
3.1 part W
3.2 Pequipment of distributed RU power consumption for remote parts 329.35 W

Table 10. Band 75 – daily energy consumption reporting

Daily energy consumption reporting

S.No Parameter Value Unit
1 Eequipment of integrated RU energy consumption Wh
3 Eequipment of distributed RU energy consumption
Eequipment of distributed RU energy consumption for the central
3.1 part Wh
3.2 Eequipment of distributed RU energy consumption for remote parts 5928.23 Wh

30
Version 1.0
Open RAN Handbook 2nd Edition
Testing energy efficiency across Open RAN systems is critical to advancing industry-wide efforts to
optimize power usage while maintaining performance. Vodafone’s current findings on the Fujitsu radio
are achieved with the activation of the Micro DTX energy-saving feature, underscoring the potential of
such innovations to reduce power consumption without compromising signal quality. To further these
insights, we will conduct comparative tests by deactivating the energy-saving feature to measure the
differences in power consumption. Additionally, we plan to simulate real network conditions in the lab
by transmitting on all three bands simultaneously, with and without Micro DTX, to evaluate its impact on
power consumption and signal integrity. Vodafone will also analyze the feature's performance under full
load conditions, using symbol-level transmission control to validate power consumption patterns. These
tests highlight Open RAN’s pivotal role in driving energy efficiency through adaptable and innovative
solutions.

Vodafone is committed to driving advancements in energy efficiency by rigorously testing all radio
products entering our lab under similar conditions. This standardized approach ensures a consistent
evaluation of energy-saving features across different vendors and configurations. By benchmarking
power consumption and performance, we aim to identify the most energy-efficient solutions, further
promoting sustainability and innovation in Open RAN networks.

6.2 O-DU / CU / Cloud (chipsets / GPUs, accelerators, CaaS)

6.2.1 Vodafone perspective
In the rapidly evolving landscape of telecommunications, energy efficiency has become a strategic
priority, especially in Open RAN networks where minimizing OpEx and supporting global sustainability
are key goals. For MNOs, optimizing energy usage within O-DU / O-CU / O-Cloud components offers the
opportunity to significantly reduce power consumption, directly impacting cost efficiency and
environmental sustainability. Open RAN modular, vendor-agnostic architecture provides an ideal
foundation for implementing advanced power-saving technologies, enabling dynamic adjustments based
on real-time traffic and operational conditions.

6.2.2 O-RAN specifications

O-RAN ALLIANCE WG6 is undertaking a technical report [8] to study several key advancements and
strategies for energy savings within the O-Cloud infrastructure. It describes O-Cloud mainstream energy
management techniques including:

• CPU energy management: Adjusts CPU frequency and voltage to reduce energy consumption.

• Clock gating: Disables parts of the CPU circuit to save power.

• Adaptive link rate: Adjusts network link data rates during low traffic periods.

• Idle and performance states: Uses different states to manage power based on usage.

The technical report also provides the following energy-saving use cases for O-Cloud:

• Node shutdown in idle mode: Shuts down idle nodes to save energy, based on predictive models
of resource usage.

31
Version 1.0
Open RAN Handbook 2nd Edition
• CPU core frequency and pinning configuration: Adjusts CPU configurations for virtualized network
functions (VNFs) / Cloud-native functions (CNFs) during low usage periods to save energy.

• C-State usage for NF deployment: Configures CPU states to reduce power consumption during NF
deployments.

• Node cluster mode selection: Selects specific modes for node clusters to achieve energy savings.

6.2.3 Different options

Several options exist for implementing energy-saving features for the O-DU / O-CU / O-Cloud through
dynamic adjustments of the processors' C-States, P-States, and Uncore Frequency. These dynamic
adjustments need to interoperate seamlessly with O-DU / O-CU workloads to ensure that customer
experiences, service level agreements, and network performance stay optimal at all times.

C-States mainly impact application-level power savings. It is possible to use dynamic adjustments of C-
State to control the sleep mode per CPU core. C-States (refer to Figure 13) usually start in C0, which is
the normal CPU operating mode. With increasing Cx, the CPU sleep mode becomes deeper, and more
circuits are turned off. However, more time is required to bring the CPU back to C0 or to wake it up.

P-States mainly impact server-level power savings. Dynamic adjustments to P-States allow each CPU core
to run at a different frequency or voltage (execution power).

Uncore Frequency refers to the clock speed of a CPU's non-core components, such as the memory
controller, cache, and other integrated peripherals. This frequency is separate from the core frequency,
which is the speed at which the CPU cores operate.

Figure 13. C-States (image provided by Intel)

6.2.4 Collaboration test setup

Through various Open RAN PlugFest demonstrations and live network validations, Vodafone has
partnered with leading vendors to explore the potential of energy usage optimization in O-DU / O-CU /
O-Cloud to achieve breakthroughs in energy savings.

32
Version 1.0
Open RAN Handbook 2nd Edition
These energy-saving initiatives with traffic profiles defined in ETSI TS 103-786 [9] and Vodafone's live
traffic profile simulated, including up to 20% energy reduction at key events like Mobile World Congress
(MWC) 2023 and O-RAN PlugFest Fall 2023. This result underscores the importance of innovations such
as dynamic C-State and P-State optimizations, real-time KPI-driven power management, and the
integration of hardware accelerators. By validating these multi-vendor solutions, Vodafone is not only
advancing industry standards for energy-efficient Open RAN deployments but also demonstrating a
practical path to sustainable network growth.

6.2.5 Gains and achievements in energy savings optimization research

The various PlugFest and MWC demonstrations reveal significant gains in energy efficiency across
multiple dimensions of O-RAN systems. We summarize the key gains and tactics in Table 11.

Table 11. Energy Savings Optimization Table

Demo Milestone PlugFest Spring PlugFest Fall 2022 MWC Barcelona 2023 PlugFest Fall
2022 2023

Energy Savings 9-12 percent (up Dynamic savings (no ≤20 percent 20 percent
(percentage) to 15 percent) exact percentage)

Main Needle C1 + P-States Closed loop C-State, P-State, HW

Mover combination Uncore Frequency Accelerator +
P-State control AAL
Interface

Test Setup Cloud-based Closed-loop control Closed-loop dynamic New

Changes multi-vendor O- based on KPIs control with Near-RT- hardware
RAN RIC and SMO (Intel 4th
Gen)

Optimization C1 state on by CPU management via Telemetry, dynamic Core

Techniques default, dynamic P-State, Uncore power control reduction, P-
P-States Frequency State tuning

Key Observations Best results with PRB, CPU Utilization A closed loop ensures HW and SW
C1 & P-States as thresholds optimal RAN abstraction
combined performance for FEC
Acceleration

Validation Keysight PA, KPI-based thresholds Live traffic + telemetry HW

Method Redfish accelerators,
DPDK BBDEV

Additional Fine-tuning C Uncore Frequency AI / ML potential Core

Energy-Saving and P-States management reduction,
Methods FEC
integration

33
Version 1.0
Open RAN Handbook 2nd Edition
Table 11 provides a clear overview of the progression of energy savings techniques from early C and P-
State optimizations to more advanced closed-loop controls and hardware acceleration in later demos.
Each stage has progressively leveraged both hardware and software innovations to push energy
efficiency in O-RAN systems.

Intel, Keysight, Radisys, Vodafone, Wind River Studio, and the University of Utah collaborated in O-RAN
Global PlugFest Spring 2022, demonstrating O-RAN energy efficiency (EE) and energy savings (ES) PoC.
Figure 14 illustrates the topology of the collaboration test setup.

Figure 14. Topology of the collaboration test setup for O-RAN PlugFest Spring 2022

Table 12 provides details on the components collaborators used at the O-RAN Global PlugFests Spring
and Fall 2022 and at MWC Barcelona in 2023.

Table 12. components collaborators at the O-RAN PlugFests Spring / Fall 2022 and at MWC 2023

Collaborators Components 2022 Spring 2023 Fall

/ Fall PlugFest
PlugFest +
MWC 2023
Intel •O-DU and O-CU hardware including servers based on Yes
3rd Generation Intel® Xeon® Scalable Processors with
telemetry and power management capabilities
•Intel® vRAN Accelerator ACC100 Adapter (DU)
O-DU and O-CU hardware including servers based on Yes
4th Generation Intel® Xeon® Scalable Processor with
Intel vRAN Boost, telemetry and power management
capabilities

•Intel® QuickAssist Adapter 8970 (CU) Yes Yes

•Intel® FlexRAN™ software implements the O-DU Lo
and O-RAN Fronthaul functionality, executing as NFs in
containers in the Wind River CaaS platform

34
Version 1.0
Open RAN Handbook 2nd Edition
Collaborators Components 2022 Spring 2023 Fall
/ Fall PlugFest
PlugFest +
MWC 2023
Keysight •O-RU(RuSIM) and UE Emulator (UeSIM) Yes Yes
•Core Emulator (CoreSIM)
•Power Analyzer (PA2203A IntegraVision)
•Demo rApp for energy savings policy decisions
•Test Automation (Pathwave Test Automation)
•Reporting and Analytics (Atlas TMC-Analytics) Yes

•Test cases (E-Plane Design Test Suite for O-DU/O-CU)

Radisys •Containerized O-DU and O-CU software automatically Yes Yes

allocates vCPU and VF resources from a resource pool
provided by Wind River CaaS Platform running on a
server with Intel Xeon processor and hardware
acceleration
Dell •Two Dell Hook (XR8620) Yes
Vodafone •Live network traffic models, KPIs and Network Yes Yes
topology
Wind River •O-Cloud using Wind River Studio Yes Yes
University of •Lab and Lab-as-a-Service Yes
Utah

6.2.6 Measurement techniques

We used a common measurement methodology across all three PlugFests (2022, 2023 and 2024).

The Keysight Energy Plane Test Suite topology, depicted in Figure 15, traverses multiple aspects with a
test automation and reporting framework following the methodology described in ETSI TS 103 786 [9].
Keysight UE / O-RU Emulator (RuSIM) and Core Emulator (CoreSIM) generate a dynamic traffic load and
channel emulation by expecting a lower traffic level to result in lower energy consumption.

Figure 15. Keysight energy plane test suite topology

35
Version 1.0
Open RAN Handbook 2nd Edition

The traffic load is modelled after Vodafone commercial network clusters. At each test iteration, DMTF
Redfish or the Keysight Power Source / Analyzer automatically evaluate the DUT's energy consumption.
The Keysight UE / O-RU emulation (RuSIM) also emulates groups of UEs with high / medium / low path
loss and different file sizes (small / medium / large) for data transfer as pictured in Figure 16 and per ETSI
TS 103 768 [9].

Figure 16. UE groups with different path loss and traffic load

In such cases, there are around 2,000 automated test iterations (refer to Figure 17) to execute to ensure
a fully comprehensive coverage:

• All protocol layers: vO-DU Lo (High-PHY), vO-DU Hi, vO-CU

• Traffic load levels (low, medium, busy hours)

• Voltage-frequency control states (P-States)

• Power consumption control states (C-States)

• Uncore Frequency scaling (UFS) technique.

Figure 17. Automated test iterations

36
Version 1.0
Open RAN Handbook 2nd Edition
The test suite can measure not only power consumption but also the subscriber’s quality of experience
(QoE). We cannot sacrifice the subscriber’s QoE to save energy consumption and need to assure
subscriber QoE in a service level agreement (SLA).

The test suite executes repeatably and consistently and is fully automated. Finally, it can generate
detailed test reports for every test iteration and the corresponding intuitive chart, as shown in Figure 18
below. A sample of the detailed test report appears in Table 13.

Table 13. Sample of the Test Report

37
Version 1.0
Open RAN Handbook 2nd Edition

Figure 18. Test report and chart

Most importantly, the test report (refer to Table 14) shows the energy-saving gains of different
technologies (C / P-State and Uncore Frequency).

Table 14. Test report with energy-saving gain

6.2.7 Findings and results

At the O-RAN Global PlugFest Spring 2022, we conducted experiments to evaluate the best approaches
to configure C and P-States to achieve the best possible energy savings gains without impacting
customers' experiences and network performances. We observed that:

1. Combining C1 and P-States (2.1GHz) power savings optimization resulted in the highest energy
savings compared to enabling them separately.

2. An additional percent energy savings can be achieved when the C1 state is enabled by default for all
load levels, as opposed to only being enabled for low-load scenarios. Switching between C0 and C1
states seems to consume additional energy, which can be saved when the CPUs are maintained in
the C1 state across all load levels.

38
Version 1.0
Open RAN Handbook 2nd Edition

At the O-RAN Global PlugFest Fall 2022, we conducted experiments to evaluate the energy savings profile
of the system under test from the closed loop automation perspective with dynamic control of the P-
States and explore additional methods to achieve greater energy savings gain.

1. Identify the KPIs for closed-loop automation of energy savings by the O-RAN xApps / rApps through
dynamic control of the P-States. We observed that the physical resource blocks (PRBs) and CPU
utilization can be used as thresholds to control the P-States for closed-loop automation.

2. Characterize the detailed timing of the closed-loop automation for each of the processing stages,
starting with the identification of the triggering event at the O-RAN xApps / rApps, the decision of
the P-State changes by the O-RAN xApps / rApps, and when the P-State change has been
implemented in the O-Cloud. We observed that the amount of energy savings gain is directly
correlated to the precision of the timing control of the P-State changes.

3. Evaluate additional energy savings gain using the Uncore Frequency adjustments after configuring
the optimal P-States. We observed that lower Uncore Frequency can lead to additional energy
savings gains (between 6.9 and 19.7 percent) on top of the energy savings gain achieved through the
optimal C and P-States settings.

At MWC Barcelona 2023, the focus was on fine-tuning the energy savings controls (C / P States and
Uncore Frequency), leveraging learnings from prior work seeking to maximize the energy savings gain
without impacting customers' experiences and network performance. We applied Vodafone’s traffic
model gathered from a live network cluster and emulated the models in the lab test environment to
simulate close-to-real-world traffic scenarios throughout the closed-loop optimization process. We
achieved an additional up to 20 percent energy savings gain, which was showcased at the MWC Barcelona
2023 event in the Vodafone booth.

In O-RAN Global PlugFest Fall 2023, we focus on characterizing the additional energy savings that can be
achieved by leveraging the state-of-the-art computing resources available at that time. Intel launched its
fourth-generation processors with integrated acceleration vRAN Boost, and no external accelerator is
required at this time. With Intel’s Gen 4 processors, all other conditions being equal, we observed:

• A system power reduction of 19.7 percent with integrated acceleration.

• A 13 percent energy savings was achieved from C1-State-enabled and uncore, as well as a P-State
frequency change to match traffic load. The O-DU / O-CU workloads have also been optimized for
energy savings.

At the O-RAN Global PlugFest Fall 2024, the focus was on estimating the energy consumption of the
individual points of delivery (PODs) within which the containerized O-DU / O-CU workloads will operate.
This will give us a more accurate characterization of the energy usage when more than one POD is
operating on a common computing infrastructure, which is highly likely in cloud deployment scenarios.

Vodafone has collaborated with its partners to explore energy consumption estimations in two Cloud as
a Service (CaaS) operating environments:

39
Version 1.0
Open RAN Handbook 2nd Edition
• Red Hat in collaboration with Dell, Keysight and OAI.

• Wind River Studio in collaboration with Dell and Intel.

The outcomes of the PlugFest Fall 2024 work provide insights into the energy consumption
instrumentation and estimations in an Open RAN environment using Red Hat Kepler (Kubernetes-based
Efficient Power Level Exporter) and Wind River Analytics (WRA) when operating in Red Hat and Wind
River Studio operating environments respectively.

It was successfully demonstrated that both methods can be used to collect energy consumption
measurements at the granularity of the POD and application resource level in run-time without degrading
application performance.

This has demonstrated to the industry that it is possible to assess power consumption across different
components and conditions within a cloud-native Open RAN deployment and ultimately help guide
strategies for power optimization in cloud-native Open RAN networks.

6.2.8 Future work and next steps

We would like to further explore the energy usage of hardware accelerators and compare the power
consumption of specialized hardware accelerators (GPUs and DPUs) versus standard CPUs when
processing RAN workloads. The goal is to determine which processing units offer the best performance
per watt in the Open RAN cloud environment, as RAN energy efficiency and savings need to be optimized
for specific deployment scenarios and real-world applications.

Furthermore, energy profile comparisons across RAN vendors and cloud providers need more work in
deployment scenarios where the Open RAN platform is deployed across multiple cloud environments.
Therefore, measurements of the energy consumption profiles across providers are needed. This can
inform decision-making about RAN software and cloud platform efficiency.

These will require continuing to refine power consumption characterization across different components
and conditions within the network and showcasing these refinements at future PlugFest events.

6.2.9 Recommendations
1. Standardization: O-RAN ALLIANCE should leverage these outcomes from the 3GPP effort to help
mobile operators save energy costs while improving subscriber QoS. With these outcomes from the
3GPP, the O-RAN ALLIANCE should continue to develop more use cases on energy efficiency and
energy savings, which will lead to evolving O-RAN specifications in various working groups. The 3GPP
R18/R19 has multiple study items working on the network energy-saving feature.

2. Operators and Adoption of AI-driven Energy Management: Operators should apply AI-based
technology to dynamically manage energy consumption, especially in multi-carrier setups where
traffic demand varies significantly.

40
Version 1.0
Open RAN Handbook 2nd Edition

6.3 RIC applications (xApps/rApps), SMO

6.3.1 Operator perspective
MNOs face an ongoing challenge to balance the growing demand for wireless services with operational
cost efficiency and environmental responsibility. Traditionally, networks are designed to operate at
maximum capacity, leading to high energy consumption, particularly during periods of low traffic
demand. With the advent of Open RAN and 5G technologies, operators can now access innovative tools
such as RAN Intelligent Controllers (RICs) and energy-saving applications that dynamically adjust network
resources, significantly improving energy efficiency without compromising the QoS.

Implementing RIC-based energy efficiency solutions in Open RAN networks represents a pivotal shift
toward smarter, more sustainable, and cost-effective network operations for operators. By dynamically
managing network resources, reducing energy consumption, and maintaining high levels of service
quality, operators can significantly reduce their OpEx while enhancing the user experience. As the
telecommunications industry continues to prioritize sustainability and cost efficiency, the adoption of
RIC-driven energy-saving use cases will become an essential strategy for operators to remain competitive
and profitable over the long term.

6.3.2 O-RAN specifications: architecture and procedure flow for cell on / off use
case
The Open RAN architecture, as defined by the O-RAN ALLIANCE, enables real-time control and
automation of RAN elements through the RIC. In the context of energy efficiency, the non-real-time (Non-
RT) RIC hosts rApps that execute long-term optimization tasks such as energy-saving algorithms, while
the near-real-time (Near-RT) RIC hosts xApps that execute time-sensitive functions, including traffic
steering. This setup is essential for the dynamic management of cells, enabling MNOs to power down
underutilized cells through the energy savings (ES) mode and redistribute traffic through active
neighboring cells.

The architecture and procedure flow for carrier / cell on / off in the context of improving the energy
efficiency of mobile networks through dynamic cell switching, with coordination between traffic steering
(TS) xApps and ES rApps, can be described as follows.

Architecture Overview

The architecture is based on the O-RAN framework, where different components communicate over
defined interfaces to perform the required functions. Figure 3 depicts the key elements involved.

The elements of the O-RAN framework include:

• Entities (or components)

o Non-RT RIC: Manages the rApps, including the Energy Saving (ES) rApp. It operates in a non-
real-time control loop, with a typical response time of one second or more, and interacts with
the Near-RT RIC through the A1 interface.

41
Version 1.0
Open RAN Handbook 2nd Edition
o Near-RT RIC: Hosts the xApps, including the Traffic Steering (TS) xApp. It operates in a near-real-
time control loop, typically ranging from 10 milliseconds to 1 second, and communicates with
the RAN via the E2 interface.

o ES-rApp: This rApp is responsible for identifying candidate cells for switch-off/switch-on based
on network traffic conditions and energy efficiency goals. It runs in the Non-RT RIC.

o TS-xApp: This xApp is responsible for offloading or onboarding users to neighboring cells when
a candidate cell is to be switched off/on. It runs in the Near-RT RIC.

o O-CU / O-DU: The components of the RAN, also known as E2 nodes that handle the radio
resources, connect to the Near-RT RIC over the E2 interface.

• Interfaces

o A1 Interface: Used for sending policy-based instructions from the Non-RT RIC to the Near-RT RIC (such as
ES-rApp instructing the TS-xApp about cell switch-off candidates).

o E2 Interface: Provides real-time control and monitoring between the Near-RT RIC and the RAN (such as
TS-xApp interacting with the O-CU/O-DU).

o O1 Interface: Plays a crucial role in the architecture by enabling the Non-RT RIC, where the ES rApp
resides, to manage and configure the RAN elements, including switching cells on or off based on AI-driven
decisions, ensuring efficient network operations and energy optimization.

o R1 Interface: It is an open logical interface within the O-RAN architecture between the rApps and the
Non-RT RIC framework. The R1 interface supports the exchange of control signaling information and the
collection and delivery of data between endpoints. The R1 interface is not utilized in the current energy-
saving use case test.

High-level Procedure Flow for Dynamic Cell Switching

The procedure flow begins with the ES-rApp detecting low traffic conditions, prompting it to prepare
specific cells for deactivation. The TS-xApp receives an alert to reroute user traffic to adjacent cells before
powering down these cells. Once the migration is complete and the target cells are empty, they can be
switched to an energy-saving mode.

6.3.3 Two options for notification from ES-rApp to TS-xApp

There are two options for notifying the TS-xApp about energy-saving actions from the ES-rApp:

• Option 1: A1 Interface (Policy-based Notification)

o The ES-rApp uses the A1 interface to send policies to the Near-RT RIC, including a “FORBID”
clause for cells that are candidates for switch-off.

o The TS-xApp receives this policy and starts the user offloading process.

• Option 2: E2 Interface (E2 Service Model, Cell Configuration and Control- E2SM-CCC)

42
Version 1.0
Open RAN Handbook 2nd Edition
o The TS-xApp uses the E2 interface to interact directly with the Near-RT RIC and the RAN, utilizing
the cell configuration and control service model (E2SM-CCC).

o This approach allows for more direct and real-time control over the cells, enabling faster
response times for cell switch-on/off actions.

6.3.4 Collaboration test setup

The test setup demonstrates a multi-vendor environment where the ES rApp and traffic steering xApp
collaborate via Juniper Networks' RIC platform, simulated with Keysight's RICtest emulation tool. The
following partners provided key components:

• Juniper Networks: Non-RT RIC and Near-RT RIC platforms

• AirHop Communications: Energy Saving rApp (ES-rApp)

• Rimedo Labs: Traffic Steering xApp (TS-xApp)

• Keysight: RICtest for RAN emulation.

Figure 19. Multi-vendor rApp-xApp coordination topology (image provided by Juniper Networks)

This logical architecture shown in Figure 19 illustrates how each component interacts through the O-RAN
interfaces, marking the specific responsibilities of each vendor. The energy-saving actions initiated by the
ES-rApp trigger procedural flows across both inner and outer loops. Although the TS-xApp can be driven
using both A1 policy-assisted and E2SM-CCC-based notifications, this particular setup demonstrates the
second option.

43
Version 1.0
Open RAN Handbook 2nd Edition

6.3.5 Keysight RICtest configuration and statistics

Keysight RICtest is a comprehensive solution for testing RIC / xApp / rApp via E2. A1 and O1 interfaces to
address customer requirements for conformance, performance, interoperability, and use case validation
tests.

In the carrier / cell on / off use case test, RICtest simulates multiple E2 nodes connecting to Non-RT RIC
via the O1 interface and Near-RT RIC via the E2 interface. These E2 nodes provide coverage and capacity
cell services to multiple groups of UEs with mobility and data traffic activities. RICtest allows designing a
cell map via a visual editor, as shown in Figure 20, and configuring the cell with rich parameters, as shown
in Figure 21.

Figure 20. Keysight RICtest map designer

44
Version 1.0
Open RAN Handbook 2nd Edition
Figure 21. Keysight RICtest cell configuration

RICtest can simulate idle capacity cell and busy coverage cell by UE behaviors to trigger energy-saving
rApp to turn off and on capacity cell / carrier. Moreover, the UEs can be handed over between the
capacity cell and the coverage cell by traffic steering xApp. All the behaviors can be observed in the
RICtest dashboard, as shown in Figure 22.

Figure 22. Keysight RICtest statistics dashboard

6.3.6 Procedural flows

The following paragraphs describe how each option operates step by step and the procedural flows for
triggering the energy-saving mechanism using the two approaches (A1 policy-based notification and
E2SM-CCC-based notification) for Traffic Steering (TS) xApp in the context of the ES use case.

45
Version 1.0
Open RAN Handbook 2nd Edition
Option 1: Traffic Steering xApp Notification through the A1 Policy (A1 Interface)

Figure 23. Option 1 topology (image provided by Rimedo Labs)

Setup Overview:

• ES-rApp: Deployed in the Non-RT RIC, responsible for deciding which cells to switch on / off based
on network KPIs.

• TS-xApp: Deployed in the Near-RT RIC, handles Traffic Steering (TS) actions like offloading users
when cells are to be switched off.

• Interfaces:

o A1 Interface: Used by the ES-rApp to notify the TS-xApp via policies.

o O1 Interface: Used by the ES-rApp to manage cell on/off operations.

o E2 Interface: Used by the TS-xApp to perform TS actions and retrieve real-time data from the
RAN.

Procedural Flow:

1. ES-rApp Decision: The ES-rApp in the Non-RT RIC continuously monitors KPIs such as user throughput,
power consumption, and PRB usage. It identifies candidate cells for energy saving and decides to switch
off certain cells.

2. A1 Policy Creation: The ES-rApp formulates an A1 policy that contains a “FORBID” clause for the selected
cells, instructing the TS-xApp that no new users should be allowed on these cells.

3. A1 Policy Notification: The is Non-RT RIC sends the A1 policy to the Near-RT RIC, notifying the TS-xApp
of the candidate cells for switch-off.

46
Version 1.0
Open RAN Handbook 2nd Edition
4. TS-xApp Action: Upon receiving the A1 policy, the TS-xApp starts offloading users from the candidate
cells. The xApp retrieves real-time data (e.g., UE information, cell load) from the RAN using the E2
interface and directs users to neighboring cells via E2SM-RC (RAN Control).

5. ES-rApp Monitoring: The ES-rApp monitors the number of RRC connections in the candidate cells. Once
the number of connections reaches zero, it confirms that all users have been successfully offloaded.

6. Cell Switch-Off: After user offloading is complete, the ES-rApp uses the O1 interface to deactivate the
cell, putting it into energy-saving mode.

7. Cell Switch-On: When traffic conditions change, the ES-rApp may decide to reactivate a cell. The ES-
rApp switches the cell on via the O1 interface and sends a request to delete the “FORBID” policy through
the A1 interface.

8. TS-xApp Action for Reactivation: Once the “FORBID” policy is deleted, the TS-xApp begins using the
reactivated cell for user traffic, ensuring that the load is balanced across the network.

Option 2: Traffic Steering xApp Notification through E2SM-CCC (E2 Interface)

Figure 24. Option 2 topology (image provided by Rimedo Labs)

Setup Overview:

• ES-rApp: Deployed in the Non-RT RIC, responsible for deciding which cells to switch on/off based
on network KPIs.

• TS-xApp: Deployed in the Near-RT RIC, handles Traffic Steering actions based on cell configuration
changes.

• Interfaces:

47
Version 1.0
Open RAN Handbook 2nd Edition
o E2 Interface: Used by the TS-xApp to receive notifications of cell configuration changes and
perform TS actions.

o O1 Interface: Used by the ES-rApp to manage cell on/off operations as well as for KPI data
collection from the RAN nodes.

Procedural Flow:

1. ES-rApp Decision: The ES-rApp in the Non-RT RIC continuously monitors KPIs such as user throughput,
power consumption, and PRB usage. It identifies candidate cells for energy saving and decides to switch
off certain cells.

2. ES-rApp Triggers Cell Configuration Change: Instead of using the A1 interface, the ES-rApp directly
triggers a cell configuration change via the E2SM-CCC service on the E2 interface. This change updates
the O-CES (O-RAN Cell Energy Saving) Management Function parameters for the affected cell.

3. TS-xApp Notification: The TS-xApp subscribes to the E2SM-CCC service and receives an indication
message whenever the O-CES Management Function attributes change. Specifically, it receives
notifications about changes to parameters such as cesSwitch, energySavingState, and
energySavingControl.

4. TS-xApp Action: When the TS-xApp detects that a cell is transitioning to the “toBeEnergySaving” state,
it begins offloading users from the cell. The xApp retrieves real-time information (such as UE IDs, and
cell load) through the E2 interface and controls handovers using E2SM-RC.

5. Cell Transition to Energy-Saving Mode: Once all users are offloaded, the cell automatically transitions
to the “isEnergySaving” state, during which most of its hardware components are deactivated, and the
cell stops serving users.

6. Cell Reactivation: When the ES-rApp decides to reactivate a cell, it changes its configuration via the
E2SM-CCC to the “toBeNotEnergySaving” state.

7. TS-xApp Action for Reactivation: The TS-xApp waits for the cell to transition to the “isNotEnergySaving”
state, indicating that it is fully reactivated. The TS-xApp then begins onboarding users back to the cell,
managing traffic across the network accordingly.

6.3.7 Findings and results

The implementation of the Energy Savings for Multi-Carrier (ESMC) rApp, which leverages AI-driven
algorithms and RAN programmability, demonstrated significant improvements in energy efficiency while
maintaining high network accessibility. The AI-based approach enabled dynamic switching of RAN
elements by differentiating between the coverage and capacity layers to ensure activation for additional
capacity only when necessary. By analysing network conditions, predicting load variations, and making
load-based decisions, the ESMC rApp optimized energy usage without compromising user Quality of
Experience (QoE).

48
Version 1.0
Open RAN Handbook 2nd Edition
The Keysight RICtest solution evaluated the ESMC rApp, simulating a representative network based on a
comprehensive Vodafone dataset; refer to Figure 25. This dataset included information from 13 sites and
41 sectors across five frequency bands, with a granularity of 15 minutes over two weeks. Key metrics
included Physical Resource Block (PRB) utilization, power consumption, user QoE satisfaction
(accessibility), the number of on-off transitions, and model complexity.

Figure 25. Number of cells across different frequency bands and sites in the Vodafone dataset

The results showed that the ESMC rApp successfully reduced energy consumption on the capacity layers
by approximately 25 percent (refer to Figure 26). It was possible to achieve this goal through the
intelligent management of capacity carriers, dynamically switching them to a low-energy state during
periods of low traffic and reactivating them as needed during high traffic. Despite these energy-saving
measures, the system maintained an exceptional accessibility level of 99.999 percent, ensuring that the
user experience was not adversely affected.

Figure 26. Traffic load, prediction, and energy-saving modes over time with AirHop’s ESMC rApp (image provided by
AirHop Communications)

49
Version 1.0
Open RAN Handbook 2nd Edition

The test is based on Vodafone dataset which applying the following O-RU characterization:

• Power is 2 x 40W

• FDD mode

• Band of 700M/800M/1800M/2100M/2600M

• 2T2R without MIMO

• Outdoor coverage.

The 25 percent energy saving gain is based on the percentage of time that the AI model decided to turn
off the capacity layers and took an average over all capacity layers.

For example: ES percentage for carriers of sector 0: [0. 0.10714286 0.33333333 0.73214286 0.98214286]
shows the energy saving percentage.

These numbers indicate the percentage of time the layers were turned off based on the AI decision. In
this example, we have 5 cells (1 coverage and 4 capacity layers).

The coverage layer is always ON, so the ES percentage is zero. The first capacity layer is OFF around 10
percent of the time, the second capacity layer 33 percent, the third capacity layer 73 percent, and the
fifth capacity layer 98 percent.

The 25 percent energy saving gain reported as a percentage number is averaged over all capacity layers
of OFF ratio decided by the AI model in the dataset.

Furthermore, the AI-driven approach efficiently balanced energy savings and performance, even during
transitions between different energy-saving modes. By minimizing the number of active capacity carriers
when network load was low and adapting as traffic increased, the rApp maximized energy efficiency while
preserving network performance.

One of the critical challenges addressed by the ESMC rApp was the coordination between multi-vendor
rApps and xApps on the RIC platform, particularly between the Energy Savings rApp and the Traffic
Steering xApp from two different providers. This coordination prevented conflicting interactions that
could degrade network performance. The cooperation between the two applications was demonstrated
during the National Telecommunications and Information Administration (NTIA) RIC Forum [9], enabled
service providers to achieve energy savings without impacting service quality. In partnership with Juniper
Networks, Vodafone, AirHop, Rimedo Labs, and Keysight, the demonstration highlighted the ability to
achieve 25% energy savings through coordinated multi-vendor rApp / xApp interactions.

6.3.8 Recommendations
1. Standardization

50
Version 1.0
Open RAN Handbook 2nd Edition
• Enhance O-RAN Specifications: The O-RAN ALLIANCE should continue refining the A1, E2, and O1
interface specifications to ensure greater compatibility between multi-vendor rApps and xApps,
enabling a more seamless integration of energy-saving and traffic-steering functions. 3GPP
R18/R19 has multiple study items working on network energy-saving features. O-RAN ALLIANCE
should create work items to leverage these outcomes from the 3GPP effort to help mobile
operators save energy costs and improve subscriber quality of service.

• Develop Common Test Methodologies: Standardized test methodology and specifications, like
what Keysight’s RICtest does, should be further developed to allow for consistent benchmarking
across different network configurations and vendor implementations.

Keysight has contributed the performance test case “Network Energy Saving with Carrier and Cell
switch off / on using Non-RT RIC” to the O-RAN ALLIANCE TIFG (Test and Integration Focus Group)
E2E (End to End) test specification.

2. Operators

• Adopt AI-driven Energy Management: Operators should be aware of the potential of AI-based
rApps to manage energy consumption dynamically, particularly in multi-carrier deployments where
traffic demand fluctuates significantly across the carrier layers.

• Leverage Open RAN Flexibility: By adopting Open RAN principles and RIC-based architectures,
operators can increase vendor diversity, which allows for more tailored and efficient network
solutions, ultimately reducing both costs and environmental impact.

This chapter highlights the technical capabilities and strategic advantages of implementing energy-
saving and traffic-steering solutions in Open RAN networks. It provides operators with a pathway
to enhance energy efficiency while maintaining network performance and user satisfaction.

51
Version 1.0
Open RAN Handbook 2nd Edition

7 Summary and Recommendations

RAN consumes more than 70 percent of the total energy consumption in MNOs’ wireless networks.
Optimization of the energy consumption and efficiency of RAN products is, therefore, a priority for the
MNOs, considering their corporate sustainability responsibilities and the topline standpoint of rising
energy costs globally (both CapEx and OpEx). Being a new entrant to the market, Open RAN needs to
demonstrate on par or better energy consumption and efficiency compared to traditional monolithic
RANs for each of the Open RAN products and in a multi-vendor Open RAN system.

Standards organizations such as the 3GPP and O-RAN ALLIANCE have been prioritizing the development
of network energy savings features in their respective standards. The O-RAN ALLIANCE standards are
complementary to the 3GPP standards: O-RAN ALLIANCE fully leverages 3GPP standardization while
specifying energy savings features for Open RAN components which are not in 3GPP’s scope, such as the
O-Cloud and RIC-enabled energy savings optimization, while ensuring that multi-vendors interoperability
aspects are well supported.

Starting in 2022, Vodafone has been exploring with its partners innovating new testing methods to
evaluate Open RAN product's energy consumption, efficiency, and savings features, as there has been
limited standardization in this space. We have shared the testing methodologies and test results for the
O-RU, O-DU / O-CU / O-Cloud, and RIC-enabled energy savings optimization used in these test campaigns
at multiple O-RAN Global PlugFests events. This handbook summarizes this information together with
the significant findings and recommendations for testing and evaluation approaches, improvements in
Open RAN products, and standardization.

The O-RU energy consumption has been evaluated using the Keysight automated Energy Plane Test Suite,
which references the ETSI-defined test methodology standardized in ETSI ES 202 706 for static energy
consumption measurement for the base station. The category A O-RU energy savings feature (micro DTX
energy savings feature, which switches on / off its power amplifier at the symbol level) has been
characterized and demonstrated significant energy savings gains compared to when this feature is not
enabled.

The O-DU / O-CU / O-Cloud energy consumption and efficiency have been evaluated for multiple
objectives with the Keysight automated Energy Plane Test Suite. These objectives include:

1. Benchmarking energy consumption and efficiency referencing the ETSI-defined test methodology
standardized in ETSI TS 103-786 for dynamic energy consumption and efficiency measurement for
the base station.

2. Exploring energy savings optimization with close to real world’s traffic patterns; this goal is achieved
by emulating Vodafone commercial network cluster of base stations traffic profiles in a lab
environment and ensuring that energy savings implementation doesn’t adversely impact customers
experiences and network performances.

3. Instrumented payloads to assist with the development and characterization of the energy profile of
the O-DU / O-CU / O-Cloud to assist with the design of RIC-enabled optimization algorithms.

4. Collection of energy consumption measurements of the cloud virtual workloads in run-time without
degrading the application performance.

52
Version 1.0
Open RAN Handbook 2nd Edition
Testing with both points 1 and 2 has demonstrated that a significant amount of energy can be saved
when the O-DU / O-CU workloads can be optimally designed to dynamically manage the processors’ C
and P-States and Uncore Frequency in response to changing workloads.

Testing with point 3 has shown early results that the multi-vendors O-DU / O-CU / O-Cloud system will
have differing energy profiles. These will influence the dynamic close loop design of the RIC-enabled
optimization models and will need to be further investigated.

Testing with point 4 has revealed the possibilities of estimating the energy consumption and efficiency
of virtual workloads (DU / CU) in a virtual / cloud environment where multi-tenancy may be expected.

RIC-enabled optimization enables real-time control and automation of RAN elements with the xApps and
rApps operating on RIC platforms. Vodafone has established a RIC-enabled optimization closed loop
testbed that requires close to real-world traffic scenarios to be emulated and performs close loop
interaction with the RIC and RIC applications in the lab environment; the target of this closed loop testbed
is to test, evaluate, and benchmark RIC energy savings applications which leverages AI-driven algorithms
and RAN programmability. It has been demonstrated that significant energy savings can be achieved with
an AI-based approach, which has enabled dynamic switching of RAN elements through sensing of
network conditions, utilizing network load predictions, and making load-based decisions usage without
compromising user QoE.

Vodafone is committed to continuing to innovate with our industry partners to develop and validate
testing and optimization methods. These methods are instrumental in driving significant improvements
in energy consumption, energy efficiency, and energy savings techniques for Open RAN products, thereby
paving the way for open and greener mobile network deployments.

53
Version 1.0
Open RAN Handbook 2nd Edition

8 References
[1] 3GPP, “Study on enhancement for data collection for NR and ENDC”, 3GPP, TR 37.817 v17.0.0 2022-
04-06
[2] 3GPP, “Study on network energy savings for NR”, 3GPP, TR 38.864 v18.1.0 2023-03-30
[3] O-RAN ALLIANCE, “O-RAN Architecture Description”, O-RAN ALLIANCE, O-RAN.WG1.O-RAN-
Architecture-Description
[4] O-RAN ALLIANCE, "O-RAN Working Group 4 (Open Fronthaul Interfaces WG) Control, User and
Synchronization Plane Specification", O-RAN ALLIANCE, O-RAN.WG4.CUS.0-R004-v16.01 (October
2024).
[5] O-RAN ALLIANCE, "O-RAN Working Group 4 (Fronthaul Working Group)", O-RAN.WG4.CONF.0-R004-
v11.00 (October 2024).
[6] ETSI, "Environmental Engineering (EE); Metrics and measurement method for energy efficiency of
wireless access network equipment; Part 1: Power consumption - static measurement method",
ETSI, ETSI ES 202 706-1 V1.7.1 (2022-08).
[7] O-RAN ALLIANCE, "O-RAN Working Group 4 (Open Fronthaul Interfaces WG) Management Plane
Specification", O-RAN.WG4.MP.0-R004-v16.01 (October 2024).
[8] O-RAN ALLIANCE, "O-RAN Working Group 6 (Cloudification and Orchestration Workgroup) Study on
O-Cloud Energy Savings", O-RAN.WG6.O-Cloud-ES-v02.00 (October 2024).
[9] ETSI, “Environmental Engineering (EE); Measurement method for energy efficiency of wireless
access network equipment Dynamic energy performance measurement method of 5G Base Station
(BS)”, ETSI TS 103 786 V1.1.1 (2020-12)
[10]RIC Forum, March 26–28, 2024 - ITS https://its.ntia.gov/research/5g/2024-ric-forum

54
Version 1.0
Open RAN Handbook 2nd Edition

9 Glossary of Terms
3GPP 3rd Generation Partnership Project

AC Alternate current

AI Artificial intelligence

BLER Block error rate

BS Base station

CaaS Cloud as a Service

CapEx Capital expenditures

CCC Cell configuration and control

CPU Central processing unit

CSI-RS Channel state information reference signal

C-Plane Control plane

DC Direct current

DPU Data processing unit

DRX Discontinuous reception

DTX Discontinuous transmission

DV Data volume

E2E End-to-end

E2SM E2 Service model

EC Energy consumption

EE Energy efficiency

eNB Evolved node B

ES Energy savings

ESF Energy saving feature

ESMC Energy savings for multi-carrier

ETSI Europe Telecommunication Standard Institute

FDD Frequency division duplexing

FFT Fast Fourier transform

55
Version 1.0
Open RAN Handbook 2nd Edition
FR Frequency range

gNB Next generation node B

GPU Graphical processing unit

iFFT Inverse fast Fourier transform

KEPLER Kubernetes-based efficient power level exporter

KPI Key performance indicator

ML Machine learning

MNO Mobile network operator

MWC Mobile World Congress

M-Plane Management plane

Near-RT RIC Near real-time RAN intelligent controller

Non-RT RIC Non-real-time RAN intelligent controller

NG-RAN Next-generation radio access network

NR New radio

NTIA National Telecommunications and Information Administration

OFDM Orthogonal frequency division multiplexing

OTA Over-the-air

O-CU Open central unit

O-DU Open distributed unit

O-RAN Open radio access network

O-RU Open radio unit

OpEx Operating expense

PA Power amplifier

PDSCH Physical downlink shared channel

POD Point of delivery

PRB Physical resource block

QoE Quality of experience

QoS Quality of service

56
Version 1.0
Open RAN Handbook 2nd Edition
RAN Radio access network

RF Radio frequency

RIC Radio intelligent controller

RMSI Remaining minimum system information

RP Release proposal

RU Radio unit

SSB Synchronization signal block

TDD Time-division duplexing

TIFG Test and integration focus group

TR Technical report

TRX Transceiver

TS Traffic steering

SCell Serving-cell

SIB1 System information block type 1

SLA Service level agreement

SMO Service management orchestration

S-Plane Synchronization plane

SSB Synchronization signal block

UE User equipment

ULPI UpLink performance improvement

UPT User perceived throughput

U-Plane User plane

WG Working group

Wh Watts per hour

57
Version 1.0

This information is subject to change without notice. © Keysight Technologies, 2025, Published in USA, March 3, 2025, 7125-1010.EN