Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

Energy-Aware CPU Orchestration in O-RAN: A dApp-Driven Lightweight Approach

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A dApp co-located with the O-DU can cut CPU power nearly in half by combining thread affinity with frequency scaling, without missing real-time deadlines.

desk verdict Promising dApp packaging for DU CPU control, but the closed-loop claim is not tested; the 49% saving is a static-configuration result. read the letter →

arxiv 2508.00629 v1 pith:O6D2MNG4 submitted 2025-08-01 cs.NI cs.OScs.PF

Francisco Crespo , Javier Villegas , Carlos Baena , Eduardo Baena , Sergio Fortes , Raquel Barco This is my paper
classification cs.NIcs.OScs.PF
keywords O-RANdAppCPUaffinityfrequencyscalingpowerconsumptionperftelemetryreal-timeschedulingsoftwarizedRAN
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that the missing piece in energy-efficient softwarized radio access is a distributed application (dApp) that lives on the same server as the baseband processing unit, reads the operating system's own performance counters, and adjusts CPU scheduling within a 1 ms transmission-time slot. Its experiments compare four CPU policy presets on an open-source 5G stack and find that pairing thread-to-core affinity with on-demand frequency scaling draws 19.5 W instead of 38 W, a 49 percent reduction, with no throughput loss. The authors claim this validates a vendor-agnostic, O-RAN-compliant control loop that does not require kernel changes or access to proprietary RAN internals. If it holds, operators could get a substantial fraction of DVFS energy savings without jeopardizing real-time deadlines.

What carries the argument

The central object is the dApp, a containerized user-space controller deployed next to the O-DU threads, which polls perf-level counters and actuates three OS knobs: thread-to-core affinity, core isolation, and CPU governor frequency scaling. The energy argument rests on the power model $P_f = P_s + k f^2$, whose quadratic dynamic term makes peak-frequency operation expensive; the paper measures this curve on its own server and adopts the 'frequency affinity' strategy of keeping voltage and frequency steady, clustering cache-friendly threads on dedicated cores, and letting the governor drop frequency only when utilization falls. The testbed design isolates the contribution of each knob through four configurations, and the dApp's rule-based controller is meant to choose the affinity-frequency combination subject to the deadline constraint $\mathrm{Latency}(\tau_j) \le T_\mathrm{TTI}$.

What would settle it

Run the dApp as a live closed loop under a time-varying load trace while recording per-slot latency and averaged wall power: if sustained power does not approach the static 19.5 W figure, or any slot exceeds the 1 ms TTI budget during a frequency or affinity transition, the paper's central claim would be refuted.

Watch

Extended reading notes

Core claim

The paper's central claim is that a lightweight user-space dApp co-located with the O-DU can orchestrate CPU affinity, core isolation, and frequency scaling from standard Linux performance-counter telemetry—context switches, instructions per cycle, and cache misses per kilo-instruction—and that this joint control works without breaking the 1 ms TTI budget. The signature result is Configuration IV, which applies static affinity plus demand-following frequency scaling on an eight-core server running an open-source 5G stack: average power falls to 19.5 W from 38 W for the performance-governor baseline, a 49 percent reduction, while downlink TCP throughput rises slightly from 50 to 51 Mbps. The same configuration reduces context switches below $1.2\times10^5$, keeps IPC above 1.5, halves MPKI, adds under 0.3 percent CPU overhead, and, the authors state, never exceeds the 1 ms slot deadline. The contribution is placed as the missing sub-millisecond layer between O-RAN's centralized controllers and the OS scheduler.

Load-bearing premise

The central assumption is that the static 19.5 W configuration represents what the closed-loop dApp sustains under changing traffic, and that its user-space affinity and frequency writes always complete within the 1 ms slot budget.

Editorial extensions

If this is right

  • If correct, an affinity-plus-frequency policy replaces the performance governor at the same throughput level: 19.5 W versus 38 W on the measured server.
  • The sub-0.3 percent overhead claim separates scheduling improvements from measurement artifacts, so the savings should transfer to other stacks that expose the same user-space knobs.
  • Staying within the 1 ms TTI budget means the control loop can act at slot level, below the tens-of-milliseconds reach of xApps and rApps, making per-TTI energy optimization feasible.
  • Because the dApp uses only standard telemetry and user-space interfaces, it is deployable on O-RAN-compliant O-DUs without vendor instrumentation or kernel modifications.
  • Reduced context switches, higher IPC, and halved MPKI indicate less scheduler noise and cache thrashing, which the paper ties directly to energy waste.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial extension: the paper compares four static presets rather than exercising the dApp's control loop under varying load, so the live-loop version of the 19.5 W result remains untested; a dynamic-load replay would close that gap.
  • If slot-level latency compliance holds during active frequency transitions, the same dApp pattern could be extended to other sub-TTI resources such as memory bandwidth or last-level cache partitioning.
  • The rule-based heuristic and the listed future work on thread clustering suggest the affinity decision could become a low-cost online classifier, trading the current reactive policy for one that anticipates thread behavior.
  • The vendor-agnostic design implies portability to other open-source 5G stacks, but the magnitude of the savings depends on their thread structure and cache behavior, so the 49 percent number should not be extrapolated without re-measurement.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes a lightweight O-RAN dApp co-located with the O-DU that reads perf-level telemetry (context switches, IPC, MPKI, utilization) and applies user-space controls (CPU affinity, core isolation, frequency scaling) to reduce CPU energy consumption while respecting TTI deadlines. A statistical power model (Eq. 1) is fitted to the testbed, a constrained optimization is formulated (Eq. 3), and a rule-based heuristic implementation is described. Experiments on an srsRAN Split 8 testbed compare four static policy configurations, reporting that affinity combined with frequency scaling draws 19.5 W versus 38 W for the performance-governor baseline (a 49% reduction) at 51 Mbps downlink throughput, with additional claims of less than 0.3% CPU overhead and slot-level latency never exceeding the 1 ms TTI budget.

Significance. If the dynamic closed-loop behavior were actually demonstrated, the paper would be a useful contribution to O-RAN dApp research by showing sub-TTI CPU actuation through standard OS interfaces and by connecting OS-level control to RIC orchestration. The noisy-neighbor profiling in Section V-A and the static-policy comparison in Section V-B are valuable empirical data points for the community, and the perf-based measurement framework is a practical contribution. The architecture is vendor-agnostic and aligned with the O-RAN dApp concept. However, the absence of a closed-loop evaluation leaves the central contribution — dynamic orchestration by the dApp — unverified, and several quantitative claims lack supporting measurement detail.

major comments (4)
  1. [Section V-B, Figure 7] The central claim of dynamic orchestration is not tested. Figure 7 reports only four static configurations (performance governor, static affinity, frequency scaling, and affinity plus frequency scaling) under identical traffic and RF conditions. No experiment runs the proposed dApp as a closed loop under varying load, no time series of control actions is shown, and no adaptation over time is exercised. The 49% power reduction is therefore a property of a static policy preset, not of the proposed controller. The abstract and conclusions describe closed-loop dynamic operation, but the evaluation does not support that description.
  2. [Section IV, 'rule-based heuristic'] The controller is specified only as a 'rule-based heuristic' with no rules, thresholds, adaptation period, or state machine. Equations (3)-(6) define a constrained optimization that is not solved, and no mapping from telemetry (context switches, IPC, MPKI) to control actions is given. Without this specification, the reader cannot assess whether the dApp makes correct real-time decisions, whether it oscillates, or whether it respects the constraints in Eq. (4). This is load-bearing for the claimed real-time feasibility of the approach.
  3. [Section V-B, overhead and latency claims] The statements that the dApp adds less than 0.3% CPU overhead and that 'slot-level latency never exceeded the 1 ms TTI budget' are asserted without measurement traces, trial counts, or confidence intervals. Figure 7 has no error bars or repetition counts. Given that Eq. (2) and Figure 2 argue that frequent frequency transitions are energy-inefficient and constrained by wake-up time, the paper should report how often frequency and affinity updates occur, how deadline compliance is measured, and how these observations reconcile with the use of a frequency-scaling governor in Configuration IV.
  4. [Section VI, Conclusions] The concluding section states that 'future affinity and isolation strategies... were not directly applied in this work,' which contradicts Section V-B, where Configuration II directly applies static affinity and Configuration IV directly applies affinity combined with frequency scaling. This internal contradiction needs to be resolved: either the reported savings come from manual static settings rather than from the dApp itself, or the conclusion is misworded. In either case, it directly affects what the paper claims to have validated.
minor comments (5)
  1. [Figure 1] The fit equation in Figure 1(a), '9.98 + 3.65(f fmin)2', appears to be missing a minus sign and differs from Eq. (1), which uses f^2 rather than (f - fmin)^2; please clarify the exact fitted model.
  2. [Section V-A, Figure 6] The text says 'up to five independent srsRAN gNB containers' were launched, but Figure 6 shows only 1 to 4 gNBs; please explain whether the fifth container was omitted from the plots and why.
  3. [Table I] The testbed CPU is listed as an Intel Core i9-14900K with eight available cores; since that processor has both P-cores and E-cores by default, please specify whether only the eight P-cores were enabled and how SMT was disabled.
  4. [References] Reference [23] lists only 'G. Jia and others' and should include the full author list; reference [8] is missing its title. Please correct these entries.
  5. [Section V-B] The phrase 'slot-level latency never exceeded the 1 ms TTI budget' is not accompanied by a description of how latency was measured, especially since the text earlier states that latency tests 'have not shown significant results'; please clarify the measurement methodology.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the 49% power saving is a direct measurement, not an output of the paper's fitted model or a self-citation chain.

full rationale

The paper's central quantitative claim is the measured 19.5 W / 51 Mbps result for Configuration IV in Figure 7, which is a direct experimental comparison of four static CPU policies under identical traffic and RF conditions. This result is not derived from the power model in Eq. (1), which is only fitted in Figure 1 as background motivation, nor from the optimization in Eq. (3), which the paper explicitly does not solve: Section IV states that 'control decisions are derived from a rule-based heuristic' and that 'the current heuristic approach demonstrates feasibility with minimal overhead.' No fitted parameter is renamed as a prediction, and no equation is defined in terms of the outcome it is said to explain. The self-citations in the related-work section ([7] and [10]) provide background on CPU allocation and QoE in vRANs but are not load-bearing for the energy-savings claim. The dApp framing relies on external references [4]–[6], and the frequency-affinity strategy cites external work [23]. The paper's main weakness is that the closed-loop dApp is not directly evaluated under dynamic load, so whether Configuration IV represents the actual dApp behavior is a validation gap rather than a circular reduction. The 49% saving is a measured static outcome, so the central claim is self-contained with respect to circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new theoretical entities; the dApp is a software component defined in prior O-RAN literature. The main borrowed axioms are the standard CPU power model, the validity of perf counters as scheduling-relevant telemetry, and the assumption that user-space CPU controls are safe within the TTI budget. The fitted power curves are free parameters, but they are descriptive and not used to derive the headline 49% saving.

free parameters (3)
  • Power model static term P_s = 9.98 W
    Fitted to the frequency versus power curve in Figure 1a following Eq. 1. Used for background motivation, not for computing the reported savings.
  • Power model coefficient k = 3.65 W/GHz^2
    Fitted to the quadratic frequency-power relationship in Figure 1a. Descriptive only.
  • Utilization-power fit coefficients = 52.4, -0.0544, 11.7
    Exponential fit to CPU utilization versus power in Figure 1b. Descriptive only, not used in the energy-savings claim.
assumptions (4)
  • domain assumption CPU power follows P(f) = P_s + k f^2 (Eq. 1).
    Adopted from ref [22] and fitted to local measurements in Figure 1a; used to motivate frequency control but not to compute the reported savings.
  • domain assumption Low-level counters such as context switches, IPC, and MPKI are reliable indicators of CPU scheduling inefficiency and energy waste.
    Stated in Section III following ref [13]; underlies the design but is not validated with a controlled prediction in this paper.
  • domain assumption User-space affinity, governor, and isolation changes can be applied within the TTI budget without missing deadlines.
    Required for the real-time safety claim. The paper cites switching costs in Figure 2 but does not show a dynamic-control experiment confirming deadline compliance.
  • domain assumption The srsRAN testbed with one UE and ZeroMQ background cells represents realistic DU operation.
    Section V-A; all central measurements come from this single lab setup with a single RF cell and synthetic load.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Energy-Aware CPU Orchestration in O-RAN: A dApp-Driven Lightweight Approach." pith.science (2026). https://pith.science/paper/O6D2MNG4

@misc{pith2026250800629,
  author       = {Pith},
  title        = {Pith review of: Energy-Aware CPU Orchestration in O-RAN: A dApp-Driven Lightweight Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/O6D2MNG4}},
  note         = {Machine review of arXiv:2508.00629}
}
read the original abstract

The transition toward softwarized Radio Access Networks (RANs), driven by the Open RAN (O-RAN) paradigm, enables flexible, vendor-neutral deployments through disaggregation and virtualization of base station functions. However, this shift introduces new challenges in managing CPU resources efficiently under strict real-time constraints. In particular, the interplay between latency-sensitive RAN workloads and general-purpose Operating System (OS) schedulers often leads to sub-optimal performance and unnecessary energy consumption. This work proposes a lightweight, programmable distributed application (dApp) deployed at the Distributed Unit (DU) level to dynamically orchestrate CPU usage. The dApp operates in closed loop with the OS, leveraging thread-level telemetry like context switches, Instructions Per Cycle (IPC), and cache metrics, to adapt CPU thread affinity, core isolation, and frequency scaling in real time. Unlike existing solutions, it requires no access to proprietary RAN software, hardware-specific features, or kernel modifications. Fully compliant with the O-RAN architecture and agnostic to the underlying RAN stack, the proposed solution introduces negligible overhead while improving energy efficiency and CPU utilization. Experimental results using a commercial-grade srsRAN deployment demonstrate consistent power savings without compromising real-time processing performance, highlighting the potential of low-latency dApps for fine-grained resource control in next-generation networks

Figures

Figures reproduced from arXiv: 2508.00629 by the authors.

Figure 2
Figure 2. Power consumed and response latency at maximum [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 1
Figure 1. Relation between power consumption against CPU [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. O-RAN architecture. • Instructions Per Cycle (IPC): Efficiency measure in￾dicating how effectively the CPU executes instructions independently of frequency. • Misses Per Kilo Instruction (MPKI): The frequency of cache misses per thousand executed instructions, rep￾resenting memory inefficiencies that negatively impact execution latency and power consumption. Based on this context, we clearly define our control probl… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: focuses on the on-node deployment. A telemetry container gathers hardware counters and scheduler statistics, exposing a local API that a control container (the dApp) polls to evaluate the constraints from Section III. The dApp then applies thread–core affinity, governo…
Figure 5
Figure 5. Figure 5: Deployment for evaluating Noisy Neighbor effects. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Metrics accumulated by the process and accumulated by the CPU pool in the srsRAN deployment. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Power consumption using core affinity and frequency control strategies on srsRAN 5G. [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Programmable and GPU-Accelerated Edge Inference for Real-Time ISAC on NVIDIA Aerial Testbed

    cs.NI 2025-12 accept novelty 7.0 of 10

    A programmable GPU-accelerated framework for real-time AI on 5G edge RAN enables ISAC applications, with cuSense demonstrating 77 cm mean localization error using existing uplink signals on a compliant deployment.

  2. Energy-Latency Trade-offs in O-RAN with Distributed Baseband Processing and AI Inference

    cs.NI 2026-08 conditional novelty 6.0 of 10

    An O-RAN model with AI inference costs shows edge processing minimizes latency at higher energy, centralized processing minimizes energy at higher latency, and an energy budget picks the mix.

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages · cited by 2 Pith papers

  1. [1]

    O-RAN: Towards an Open and Smart RAN,

    O.-R. Alliance, “O-RAN: Towards an Open and Smart RAN,” Online, 2024, available: https://www.o-ran.org, Accessed: Jan. 29, 2024

    work page 2024

  2. [2]

    O-RAN: Disrupting the Vir- tualized RAN Ecosystem,

    A. Garcia-Saavedra and X. Costa-P ´erez, “O-RAN: Disrupting the Vir- tualized RAN Ecosystem,” IEEE Communications Standards Magazine, vol. 5, no. 4, pp. 96–103, 2021

    work page 2021

  3. [3]

    5GPerf: Profiling Open Source 5G RAN Components Under Different Architectural Deployments,

    C. Wei, A. Kak, N. Choi, and T. Wood, “5GPerf: Profiling Open Source 5G RAN Components Under Different Architectural Deployments,” in Proc. ACM SIGCOMM Workshop 5G-MeMU , 2022, pp. 43–49

    work page 2022

  4. [4]

    dApps: Distributed Applications for Real-Time Inference and Control in O- RAN,

    S. D’Oro, M. Polese, L. Bonati, H. Cheng, and T. Melodia, “dApps: Distributed Applications for Real-Time Inference and Control in O- RAN,” IEEE Communications Magazine , vol. 60, no. 11, pp. 52–58, 2022

    work page 2022

  5. [5]

    dApps for Real-Time RAN Control: Use Cases and Requirements,

    O-RAN Alliance, “dApps for Real-Time RAN Control: Use Cases and Requirements,” Online, 2024, available: https://mediastorage.o-ran.org/ngrg-rr/nGRG-RR-2024-10-dApp% 20use%20cases%20and%20requirements.pdf, Accessed: Jan. 29, 2024

    work page 2024

  6. [6]

    Lacava, L

    A. Lacava, L. Bonati, N. Mohamadi, R. Gangula, F. Kaltenberger, P. Johari, S. D’Oro, F. Cuomo, M. Polese and T. Melodia, dApps: En- abling Real-Time AI-Based Open RAN Control, 2025, arXiv:2501.16502 [cs.NI]

    work page arXiv 2025

  7. [7]

    Impact of CPU Resource Allocation on vRAN Per- formance in O-Cloud,

    M. Herv ´as-Guti´errez, E. Baena, C. Baena, J. Villegas, R. Barco, and S. Fortes, “Impact of CPU Resource Allocation on vRAN Per- formance in O-Cloud,” [Online]. Available: https://doi.org/10.36227/ techrxiv.23792580.v1, 2023

    work page

  8. [8]

    Implementation of a 4G/5G Base Station Using the srsRAN Software and the USRP Software Radio Module,

    W. Flakowski, M. Krasicki, and R. Krenz, “Implementation of a 4G/5G Base Station Using the srsRAN Software and the USRP Software Radio Module,” in J. Telecommun. Inf. Technol., vol. 3, no. 3, 2023, pp. 30–40

    work page 2023

Show all 25 references
  1. [9]

    Amarisoft: 4G/5G Software Solutions,

    Amarisoft, “Amarisoft: 4G/5G Software Solutions,” Online, available: https://www.amarisoft.com, Accessed: Feb. 11, 2025

    2025

  2. [10]

    Assessing the Impact of Computational Resources to the Quality of Experience Provided by vRANs,

    C. Baena, M. Herv ´as-Guti´errez, E. Baena, J. Villegas, R. Barco, and S. Fortes, “Assessing the Impact of Computational Resources to the Quality of Experience Provided by vRANs,” vol. 11, pp. 102 944– 102 948, 2023

    2023

  3. [11]

    Characterizing the Computational and Memory Requirements of Virtual RANs,

    S. Pramanik and A. Ksentini and C. Chiasserini, “Characterizing the Computational and Memory Requirements of Virtual RANs,” in Pro- ceedings of WONS , 2022, pp. 1–8

    2022

  4. [12]

    Estimating VNF resource requirements using machine learning techniques,

    H. Jmila, M. I. Khedher, and M. A. El Yacoubi, “Estimating VNF resource requirements using machine learning techniques,” in Proc. 24th Int. Conf. Neural Inf. Process. (ICONIP). Guangzhou, China: Springer, Nov. 2017, pp. 883–892

    2017

  5. [13]

    AIRIC: Orchestration of Virtualized Radio Access Networks With Noisy Neigh- bours,

    J. X. S. Lozano, A. Garcia-Saavedra, X. Li, and X. C. Perez, “AIRIC: Orchestration of Virtualized Radio Access Networks With Noisy Neigh- bours,” IEEE Journal on Selected Areas in Communications, vol. 42, pp. 432–445, 2 2024

    2024

  6. [14]

    [Online]

    Amarisoft, Linux Setup for Best Performance , 2024, accessed: 2025- 03-10. [Online]. Available: https://tech-academy.amarisoft.com/lteenb. doc#Linux-setup-for-best-performance-1

    2024

  7. [15]

    srsRAN Documentation, Release 23.11,

    S. R. Systems, “srsRAN Documentation, Release 23.11,” Online, [On- line]. Available: https://docs.srsran.com/projects/4g/en/latest/, Accessed: Mar. 12, 2024

    2024

  8. [16]

    OpenAirInterface: Democratizing innovation in the 5G Era,

    F. Kaltenberger, A. P. Silva, A. Gosain, L. Wang, and T.-T. Nguyen, “OpenAirInterface: Democratizing innovation in the 5G Era,” Computer Networks , vol. 176, p. 107284, 2020. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S1389128619314410

    2020

  9. [17]

    Towards Energy Efficient 5G vRAN Servers,

    A. Kalia, N. Lazarev, L. Xue, X. Foukas, B. Radunovic, and F. Y . Yan, “Towards Energy Efficient 5G vRAN Servers,” in Proceedings of the 22nd USENIX Symposium on Networked Systems Design and Implementation (NSDI ’25) , April 2025. [Online]. Available: https://www.usenix.org/co...

    2025

  10. [18]

    Improving Energy Ef- ficiency In Open RAN Through Dynamic CPU Scheduling,

    S. Urumkar, B. Ramamurthy, and S. Sharma, “Improving Energy Ef- ficiency In Open RAN Through Dynamic CPU Scheduling,” in 2023 IEEE International Conference on Advanced Networks and Telecommu- nications Systems (ANTS) , 2023, pp. 288–293

    2023

  11. [19]

    Energy Efficiency And Fault Tolerance In Open RAN And Future Internet,

    S. Urumkar and B. Ramamurthy and S. Sharma, “Energy Efficiency And Fault Tolerance In Open RAN And Future Internet,” in 2023 IEEE In- ternational Conference on Advanced Networks and Telecommunications Systems (ANTS), 2023, pp. 1–3

    2023

  12. [20]

    Energy-Efficient and Delay-Guaranteed Joint Resource Allocation and DU Selection in O-RAN,

    T. Pamuklu, S. Mollahasani, and M. Erol-Kantarci, “Energy-Efficient and Delay-Guaranteed Joint Resource Allocation and DU Selection in O-RAN,” in 2021 IEEE 4th 5G World Forum (5GWF) , 2021, pp. 99– 104

    2021

  13. [21]

    Enhancing Energy Efficiency in O-RAN Through Intelligent xApps Deployment,

    X. Liang, A. Al-Tahmeesschi, Q. Wang, S. Chetty, C. Sun, and H. Ah- madi, “Enhancing Energy Efficiency in O-RAN Through Intelligent xApps Deployment,” 07 2024, pp. 1–6

    2024

  14. [22]

    A Stochastic Model for Estimating the Power Consumption of a Processor,

    W. Dargie, “A Stochastic Model for Estimating the Power Consumption of a Processor,” IEEE Transactions on Computers , vol. 64, 04 2014

    2014

  15. [23]

    Frequency Affinity: Analyzing and Maximizing Power Efficiency in Multi-core Systems,

    G. Jia and others, “Frequency Affinity: Analyzing and Maximizing Power Efficiency in Multi-core Systems,” in Proceedings of IEEE MASCOTS, 2012, pp. 495–497

    2012

  16. [24]

    USRP N300 - Ettus Research,

    Ettus Research, “USRP N300 - Ettus Research,” 2025, accessed: 2025-03-20. [Online]. Available: https://www.ettus.com/all-products/ usrp-n300/

    2025

  17. [25]

    [Online]

    SIMCom. [Online]. Available: https://www.simcom.com/product/ SIM8380G-M2.html 9 Francisco Crespo received his degree in Telecommunication Technologies Engineering at the University of Malaga, Spain, in 2022. He works as a researcher at the University of M ´alaga, focusing on t...

    2010

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.