pith. sign in

arxiv: 2604.24465 · v1 · submitted 2026-04-27 · 💻 cs.NI

Large-scale wireless network management via Open-RAN Tandem Apps: Cell on/off switching use case

Pawe{\l} Kryszkiewicz , {\L}ukasz Ku{\l}acz , Marcin Paku{\l}a , Marcin Dryjanski , Marcin Hoffmann , Piotr Skrzypczak , Heiko Lehmann , Martin Stahn This is my paper

Pith reviewed 2026-05-08 01:17 UTC · model grok-4.3

classification 💻 cs.NI
keywords O-RANTandem Appscell on/off switchingnetwork managementRAN Intelligent Controllersheterogeneous networksreal data evaluationoptimization
0
0 comments X

The pith

Tandem Apps split optimization between O-RAN's near-RT and non-RT controllers to deliver low-complexity cell management with a global network view.

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

The paper introduces Tandem Apps as paired optimization mechanisms that run across the two tiers of O-RAN controllers to handle rising mobile-network complexity. Architectural and functional splitting between the controllers produces an agile solution with limited control overhead that still supports decisions informed by the full network state. The concept is shown in practice through a cell on/off switching implementation tested on a large heterogeneous network using real data and confirmed to meet O-RAN standards.

Core claim

Tandem Apps consist of a pair of tightly coupled optimization mechanisms running on both the near-real-time and non-real-time RAN Intelligent Controllers, designed via architectural and functional splitting to achieve an agile, low-complexity solution that preserves a global network view, as demonstrated by their implementation and evaluation for cell on/off switching in a large heterogeneous network with real network data.

What carries the argument

Tandem Apps: a pair of tightly coupled optimization mechanisms on the near-RT and non-RT RICs created through architectural and functional splitting between the two controllers.

If this is right

  • Enables practical management of growing mobile-network complexity without the high overhead of full centralization or the sub-optimality of purely distributed methods.
  • Supports efficient cell on/off decisions in large heterogeneous networks when evaluated against real data.
  • Remains fully compliant with the existing O-RAN standard and can run on commercial network software.
  • Offers a template for other optimization tasks that benefit from hierarchical control across the two RIC tiers.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same splitting pattern could be applied to related problems such as dynamic resource allocation or traffic steering to reduce energy consumption in dense 5G deployments.
  • If the coupling proves stable, Tandem Apps may influence the design of hybrid control loops in future wireless standards beyond O-RAN.
  • Testing the approach under varying traffic loads or with imperfect measurement data would reveal how robust the global-view benefit remains in realistic conditions.

Load-bearing premise

The chosen split between the two RICs must deliver performance close enough to a centralized optimum while keeping control overhead and complexity low, without hidden costs from the coupling mechanism.

What would settle it

A direct comparison in the same large heterogeneous network showing that Tandem Apps produce substantially lower throughput, higher energy use, or markedly higher control overhead than a fully centralized optimizer would falsify the central claim.

Figures

Figures reproduced from arXiv: 2604.24465 by Heiko Lehmann, {\L}ukasz Ku{\l}acz, Marcin Dryjanski, Marcin Hoffmann, Marcin Paku{\l}a, Martin Stahn, Pawe{\l} Kryszkiewicz, Piotr Skrzypczak.

Figure 1
Figure 1. Figure 1: Tandem Apps follow O‑RAN specifica￾tions: the rApp on the Non‑RT RIC sets policies, gathers analytics, and can train AI models using large datasets. Operating on timescales of tens of seconds to minutes, it captures long‑term trends[10] and provides global, potentially com￾plex optimization over large network areas. It sup￾plies master parameters, enrichment information, or AI models to the xApp and govern… view at source ↗
read the original abstract

With growing mobile-network complexity, management and optimization have become increasingly difficult. Centralized algorithms face high control-data overhead and computational load, while distributed approaches often perform far from optimally. The O-RAN architecture introduces two tiers of RAN Intelligent Controllers (RICs), enabling hierarchical network-management schemes. This work proposes Tandem Apps: a pair of tightly coupled optimization mechanisms running on both controllers. We show how to design Tandem Apps through architectural and functional splitting to achieve an agile, low-complexity solution that still preserves a global network view. As an example, we implement Tandem Apps for cell on/off switching and evaluate them in a large heterogeneous network using real network data. Although the Tandem Apps concept is new, it remains fully compliant with the O-RAN standard, as validated using commercial network software.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes Tandem Apps, a pair of tightly coupled optimization mechanisms running across the near-RT and non-RT RAN Intelligent Controllers (RICs) in the O-RAN architecture. It describes how to achieve this via architectural and functional splitting to obtain an agile, low-complexity solution that still maintains a global network view. The approach is demonstrated on a cell on/off switching use case, evaluated in a large heterogeneous network with real network data, and asserted to remain fully compliant with the O-RAN standard.

Significance. If the evaluation substantiates performance close to a centralized optimum with low coupling overhead, the work could provide a practical path for scalable, hierarchical network management in O-RAN deployments, addressing the tension between centralized optimality and distributed feasibility in large-scale wireless systems. The use of real heterogeneous-network traces and explicit standard compliance are strengths that support deployability.

major comments (2)
  1. [Evaluation] Evaluation section: the central claim that Tandem Apps preserve a global network view while keeping overhead and complexity low is not supported by any reported comparison against a centralized benchmark (e.g., an ILP or oracle solver) on the identical real traces. Without this, it is impossible to quantify how close the split solution comes to the global optimum or to rule out hidden costs from the RIC-to-RIC coupling.
  2. [Abstract] Abstract and evaluation description: no quantitative performance numbers, baselines, control-message volumes, or latency measurements are provided, despite the assertion of evaluation on real data. This absence makes the performance and overhead claims unverifiable and load-bearing for the paper's contribution.
minor comments (1)
  1. [Introduction] A diagram illustrating the architectural and functional split between the two RICs and the coupling mechanism would improve clarity of the Tandem Apps design.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive comments, which help clarify the evaluation requirements. We address each major comment point by point below, indicating revisions where appropriate.

read point-by-point responses

  1. Referee: [Evaluation] Evaluation section: the central claim that Tandem Apps preserve a global network view while keeping overhead and complexity low is not supported by any reported comparison against a centralized benchmark (e.g., an ILP or oracle solver) on the identical real traces. Without this, it is impossible to quantify how close the split solution comes to the global optimum or to rule out hidden costs from the RIC-to-RIC coupling.

    Authors: We agree that a direct comparison to a centralized benchmark on the same traces would strengthen the claims. However, for the large heterogeneous network size and real traces used, solving an equivalent ILP is computationally intractable (requiring prohibitive runtime even with commercial solvers). This intractability is a core motivation for the tandem hierarchical design. In the revised manuscript we will add an explicit discussion of this limitation together with results on a reduced-scale sub-network where a centralized solver remains feasible, and we will report the measured RIC-to-RIC control-message volumes and latencies. revision: partial

  2. Referee: [Abstract] Abstract and evaluation description: no quantitative performance numbers, baselines, control-message volumes, or latency measurements are provided, despite the assertion of evaluation on real data. This absence makes the performance and overhead claims unverifiable and load-bearing for the paper's contribution.

    Authors: The evaluation section already contains quantitative results obtained from the real heterogeneous-network traces, including energy-efficiency gains and switching behavior relative to a static baseline. We acknowledge that these metrics are not summarized in the abstract and that overhead figures should be highlighted. In the revision we will update the abstract to include the key quantitative indicators, baselines, and coupling overhead values from the evaluation. revision: yes

standing simulated objections not resolved
  • A centralized ILP (or oracle) benchmark on the identical full-scale real traces cannot be provided, as the optimization problem is computationally intractable at that network size.

Circularity Check

0 steps flagged

Architectural proposal with no load-bearing derivations or self-referential reductions

full rationale

The paper describes a new architectural concept (Tandem Apps) for hierarchical O-RAN management via functional splitting between RICs, illustrated by a cell on/off switching implementation evaluated on real heterogeneous-network traces. No equations, optimization formulations, fitted parameters, or predictions appear in the abstract or described content that could reduce to inputs by construction. Claims of preserving global view with low overhead rest on design description and compliance validation rather than any self-definitional loop, fitted-input prediction, or self-citation chain. This is a standard self-contained engineering proposal.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities beyond the high-level Tandem Apps construct itself.

invented entities (1)
  • Tandem Apps no independent evidence
    purpose: Tightly coupled cross-tier optimization mechanism for O-RAN
    New proposed pair of applications running on both RIC tiers

pith-pipeline@v0.9.0 · 5474 in / 1104 out tokens · 70937 ms · 2026-05-08T01:17:26.395240+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

20 extracted references · 20 canonical work pages

  1. [1]

    If the average system outage βsys exceeds the outage target range, indicating too few active cells, and PP is one , signaling a ping‑pong effect , we first decrease the αoff to reduce the intensity of cell switch‑offs

    work page

  2. [2]

    Otherwise, if the average system outage βsys is greater than the target outage range, indicating too few active cells, and some cells are still off, αon is decreased to increase cell switching on intensity

    work page

  3. [3]

    Otherwise, if the average system outage βsys is below the target outage range, meaning more cells could be turned off, and the flag PP is one, indicat- ing excessive PP, the cell activation threshold αon is increased to lower cell switching on intensity

    work page

  4. [4]

    Otherwise, if the average system outage βsys is lower than the allowed value range, and there are cells that could be switched off , increment αoff to increase cell switching off intensity

    work page

  5. [5]

    learning

    otherwise, keep αon and αoff unchanged. The TS-xApp helps COOS -Tandem Apps in the operation. The TS-xApp subscribes to cell state changes. Before the cell turn -off happens, the RAN notifies the TS-xApp to take care of cleaning up a given cell (i.e., moves users out of this cell by invoking handover commands). Once empty, RAN switches off the cell and no...

    work page 2026

  6. [6]

    On the Road to 6G: Vi- sions, Requirements, Key Technologies, and Testbeds,

    C. X. Wang , et al., "On the Road to 6G: Vi- sions, Requirements, Key Technologies, and Testbeds," IEEE Communications Surveys & Tu- torials, vol. 25, no. 2, pp. 905-974, Q2 2023

    work page 2023

  7. [7]

    Understanding O -RAN: Architecture, Interfaces, Algorithms, Security, and Research Challenges,

    M. Polese, L. Bonati, S. D’Oro, S. Basagni, T. Melodia, “Understanding O -RAN: Architecture, Interfaces, Algorithms, Security, and Research Challenges,” IEEE Communications Surveys & Tutorials, 2023, Vol. 25, No. 2, pp. 1376-1411

    work page 2023

  8. [8]

    Open RAN xApps De- sign and Evaluation: Lessons Learnt and Identi- fied Challenges,

    M. Hoffmann, et al., "Open RAN xApps De- sign and Evaluation: Lessons Learnt and Identi- fied Challenges," IEEE J . on Selected Areas in Comm., vol. 42, no. 2, pp. 473-486, Feb. 2024,

    work page 2024

  9. [9]

    Toward En- ergy Efficient RAN: From Industry Standards to Trending Practice,

    L. Kundu, X. Lin , R. Gadiyar, "Toward En- ergy Efficient RAN: From Industry Standards to Trending Practice," IEEE Wireless Communica- tions, vol. 32, no. 1, pp. 36-43, Feb. 2025

    work page 2025

  10. [10]

    Energy-Efficient Base-Stations Sleep-Mode Techniques in Green Cellular Net- works: A Survey,

    J. Wu, et al., “Energy-Efficient Base-Stations Sleep-Mode Techniques in Green Cellular Net- works: A Survey,” IEEE Communications Sur- veys & Tutorials, Q2 2015,

    work page 2015

  11. [11]

    Dy- namic Base Station Switching -On/Off Strategies for Green Cellular Networks,

    E. Oh, K. Son and B. Krishnamachari, "Dy- namic Base Station Switching -On/Off Strategies for Green Cellular Networks," IEEE Trans . on Wireless Comm., vol. 12, no. 5, May 2013

    work page 2013

  12. [12]

    Enhancing Energy Effi- ciency in O-RAN Through Intelligent xApps De- ployment,

    X. Liang, et al. , "Enhancing Energy Effi- ciency in O-RAN Through Intelligent xApps De- ployment," WINCOM, Leeds, UK, 2024

    work page 2024

  13. [13]

    Design and Evaluation of Deep Reinforcement Learning for Energy Sav- ing in Open RAN

    Bordin, Matteo, et al. "Design and Evaluation of Deep Reinforcement Learning for Energy Sav- ing in Open RAN." IEEE CCNC 2025

    work page 2025

  14. [14]

    Foundations of User -Centric Cell -Free Massive MIMO

    Demir ȔT, BjɆrnson E, Sanguinetti L (2021), "Foundations of User -Centric Cell -Free Massive MIMO". Foundations and Trends in Signal Pro- cessing, Vol. 14 No. 3-4 pp. 162–472

    work page 2021

  15. [15]

    Towards Autonomous Open Radio Access Networks

    A. Kliks, et al., “Towards Autonomous Open Radio Access Networks ”, ITU Journal on Future and Evolving Technologies, June 2023

    work page 2023

  16. [16]

    Conflict mitigation framework and conflict detection in O-RAN Near- RT RIC

    C. Adamczyk, A. Kliks, "Conflict mitigation framework and conflict detection in O-RAN Near- RT RIC." IEEE Comm. Magazine 2023

    work page 2023

  17. [17]

    Spatial Traffic Distributions for Cellular Networks with Time Varying Usage Intensities Per Land -Use Class,

    D. Rose, J. Baumgarten , T. Kurner, "Spatial Traffic Distributions for Cellular Networks with Time Varying Usage Intensities Per Land -Use Class," IEEE VTC, Vancouver, Canada, 2014

    work page 2014

  18. [18]

    Multi‑scale hierar- chical rApp‑xApp tandem for energy saving using real mobile network data

    Rimedo Labs & T‑Labs. “Multi‑scale hierar- chical rApp‑xApp tandem for energy saving using real mobile network data.” MWC 2025. https://ti- nyurl.com/zj9w4c7t

    work page 2025

  19. [19]

    Two -dimensional shadow fading modeling on system level,

    C. Zhang, et al. , "Two -dimensional shadow fading modeling on system level," IEEE PIMRC, Sydney, Australia, 2012

    work page 2012

  20. [20]

    A Power Consump- tion Model and Energy Saving Techniques for 5G- Advanced Base Stations,

    M. Oikonomakou, et al., "A Power Consump- tion Model and Energy Saving Techniques for 5G- Advanced Base Stations," IEEE ICC Workshops, Rome, Italy, 2023. Pawel Kryszkiewicz is an Associate Professor at the Institute of Radiocommunications, PUT , and the Technical Director of R imedo Labs. He is the author of over 100 publications on multicarrier systems, ...

    work page 2023