Pith. sign in

REVIEW 3 minor 37 references

Leveraging Multi-Step Traffic Forecasts for Multi-Period Planning Optical Networks

T0 review · 0 major / 3 minor · reviewed 2026-06-29 · grok-4.3

Pith's one-line read Multi-step traffic predictions enable multi-period planning in reconfigurable optical networks to achieve spectrum savings while maintaining QoS.

desk verdict The paper builds a pipeline from multi-step DL traffic forecasts on real traces into ILP and heuristic multi-period planners for optical networks, with ILP winning on quality but heuristics winning on speed, and horizon length affecting provisioning errors. read the letter →

arxiv 2605.25573 v1 pith:ENL4IEI2 submitted 2026-05-25 cs.NI

classification cs.NI
keywords multi-steptrafficpredictionmulti-periodplanningopticalnetworksspectrumefficiencyencoder-decodermodelintegerlinearprogrammingservicedisruptionsqualityof
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

This paper shows how multi-step traffic forecasts, generated by an encoder-decoder model from real traces, can be fed into optimization models for planning network configurations over multiple future periods. By doing so, the network can be reconfigured proactively to match expected traffic variations, saving spectrum resources compared to static or single-period approaches. The work balances the benefits of these savings against the risk of service disruptions from too-frequent changes, using the prediction horizon as the control knob. An integer linear program finds the best plans, while faster heuristics come close. The results indicate that the right horizon depends on whether the operator prioritizes efficiency or strict service agreements.

What carries the argument

An encoder-decoder deep learning model for generating multi-step traffic predictions, integrated with an Integer Linear Programming (ILP) formulation and heuristic algorithms for multi-period network reconfiguration planning.

What would settle it

Running the planning on real traffic data and finding that the actual disruptions or spectrum usage exceed those predicted by the model under the chosen horizons would falsify the benefit claim.

Watch

Extended reading notes

Core claim

The paper establishes that leveraging multi-step-ahead traffic predictions allows an ILP formulation and heuristic algorithms to proactively optimize future network configurations in reconfigurable optical networks. This leads to enhanced spectrum efficiency while minimizing service disruptions, with the ILP providing superior solutions and the heuristics offering near-optimal performance at much lower computational times. The effect of different prediction horizons is shown to influence the trade-off between disruptions and over- or under-provisioning.

Load-bearing premise

The traffic predictions produced by the encoder-decoder model are accurate enough on real traces to support proactive optimizations that improve spectrum use without excessive provisioning errors or disruptions.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 3 minor

Summary. The manuscript proposes a framework that uses multi-step traffic forecasts generated by an encoder-decoder deep learning model trained on real traffic traces to support multi-period planning in reconfigurable optical networks. An ILP formulation and heuristic algorithms are developed to proactively optimize future configurations, trading off spectrum efficiency against reconfiguration-induced service disruptions while maintaining QoS. Evaluations under different scenarios indicate that the ILP yields superior solutions overall, with heuristics providing close-to-optimal performance at much lower runtimes; the work also examines how prediction horizon length affects over/under-provisioning and disruptions.

Significance. If the quantitative results hold, the integration of DL-based multi-step forecasting with exact and heuristic multi-period optimization offers a practical route to spectrum savings in dynamic optical networks without excessive operational disruption. The use of real traces, explicit horizon analysis, and ILP-heuristic comparison are strengths that could inform operator planning tools.

minor comments (3)
  1. [Abstract] Abstract: the statement that 'the ILP achieving better solutions overall' is presented without any numerical deltas, spectrum savings percentages, or disruption counts; adding one or two key quantitative results would make the central claim immediately assessable.
  2. [Methods / Prediction Model] The description of the encoder-decoder model and its training on real traces would benefit from explicit reporting of prediction accuracy metrics (e.g., MAE or RMSE per horizon) and how these feed into the provisioning error bounds used by the ILP.
  3. [Results] Results discussion of horizon effects on over- and under-provisioning should include a table or figure with concrete values for different horizons rather than qualitative statements only.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive evaluation of our work and the recommendation for minor revision. The summary accurately captures the contributions of the manuscript. No major comments were raised in the report.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity

full rationale

The described framework is a linear pipeline: an encoder-decoder DL model produces multi-step forecasts from external real traffic traces; those forecasts are then fed as inputs to a separate ILP/heuristic planner that trades spectrum use against reconfiguration cost. No equation, fitted parameter, or self-citation is shown to make the claimed spectrum savings or QoS guarantee equivalent to the input predictions by construction. The derivation chain therefore remains self-contained against external benchmarks.

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

Abstract-only review supplies no equations, parameters, or modeling assumptions; free_parameters, axioms, and invented_entities cannot be identified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Leveraging Multi-Step Traffic Forecasts for Multi-Period Planning Optical Networks." pith.science (2026). https://pith.science/paper/ENL4IEI2

@misc{pith2026260525573,
  author       = {Pith},
  title        = {Pith review of: Leveraging Multi-Step Traffic Forecasts for Multi-Period Planning Optical Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ENL4IEI2}},
  note         = {Machine review of arXiv:2605.25573}
}
read the original abstract

In this work, multi-step traffic predictions are leveraged to enable multi-period planning in reconfigurable optical networks. The proposed framework aims to achieve spectrum savings by adapting the network to predicted time-varying conditions while ensuring the necessary quality-of-service (QoS) levels. Since frequent network (re)configurations may lead to undesired service disruptions, traffic predictions spanning various prediction horizons are exploited to balance the trade-off between spectrum savings and service disruptions. For multi-step-ahead prediction, an encoder-decoder deep learning model is employed to analyze real traffic traces. Subsequently, an Integer Linear Programming (ILP) formulation and heuristic algorithms are developed that use the predictions to proactively (re)optimize future network configurations, enhancing spectrum efficiency while minimizing service disruptions. The approaches are utilized under different scenarios, with the ILP achieving better solutions overall, and the heuristics achieving solutions close to the ILP at significantly lower running times. Further, the results present the effect of the prediction horizon on disruptions and over- and under- provisioning, showcasing that the prediction horizon selection greatly depends on the network operator targets in both network performance and predefined service level agreements.

Figures

Figures reproduced from arXiv: 2605.25573 by the authors.

Figure 1
Figure 1. A generic ED-LSTM architecture. ∀t ′ = t + 1, ..., t + u, where g(·) is the activation function of the output layer. As the ED-LSTM is trained as a regressor, the de￾coder’s output is connected with a fully connected dense layer. The Adam algorithm is utilized for model optimization over dataset D={xt , yt} n t=1 to reduce the mean squared error (MSE) loss function, where n denotes the observed traffic sequences. Af… view at source ↗
Figure 2
Figure 2. ED-LSTM training evolution vs. MSE loss for D1 (node 8 - u=2), D2 (node 2 - u=4), and D3 (node 5 - u=6) Abilene nodes. Sample Id 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Bit-rate (Gbps) 0 20 40 60 80 100 120 Predictions Fluctuations Ground Truth Sample Id 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Bit-rate (Gbps) 20 40 60 80 100 120 140 Predictions Fluctuations Ground Truth Sample Id 0 500 1000 150… view at source ↗
Figure 3
Figure 3. Testing-Prediction Samples for D1 (node 8 - u=2), D2 (node 2 - u=4), and D3 (node 5 - u=6) Abilene nodes. horizon increases, model accuracy drops. As degraded model accuracy is expected to affect service level agreements (SLAs) in service provisioning (i.e., violations in targeted QoS) the predic￾tion horizon must be carefully selected. Overall, in this work, models capable to predict the traffic demand with suffici… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: MMD-SA and MAD-SA examples given the bit-rates predicted for each time interval t + i and for each connection ci , with u=4. demands to be established in the network are sorted in descend￾ing order, based on their bit-rates selected at each planning in￾terval (i.e., ea…
Figure 5
Figure 5. Figure 5: Disruption, Over-provisioning, and Under-provisioning versus the number of multi-step ahead predictions (u). seen from [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 1 canonical work pages

  1. [1]

    Ericsson Mobility Report: Early movers pursue performance-based business models,

    Ericsson, “Ericsson Mobility Report: Early movers pursue performance-based business models, ” https://www.ericsson. com/en/reports-and-papers/mobility-report/dataforecasts/ mobile-traffic-forecast (2024). Accessed: 2026-02-10

  2. [2]

    HuaweiT ech Issue 096: Reshaping the future with AI-driven traffic growth,

    Huawei, “HuaweiT ech Issue 096: Reshaping the future with AI-driven traffic growth, ” https://www.huawei.com/en/publications/huaweitech/ 202401 (2024). Accessed: 2026-02-12

  3. [3]

    Survey on machine learning for traffic-driven service provisioning in optical networks,

    T . Panayiotou, M. Michalopoulou, and G. Ellinas, “Survey on machine learning for traffic-driven service provisioning in optical networks, ” IEEE Commun. Surv. & T utorials 25, 1412–1443 (2023)

  4. [4]

    AI/ML-as-a- Service for optical network automation: Use cases and challenges [in- vited],

    C. Natalino, A. Panahi, N. Mohammadiha, and P . Monti, “AI/ML-as-a- Service for optical network automation: Use cases and challenges [in- vited], ” IEEE/OPTICA J. Opt. Commun. Netw. 16, A169–A179 (2024)

  5. [5]

    A data- driven bandwidth allocation framework with QoS considerations for EONs,

    T . Panayiotou, K. Manousakis, S. Chatzis, and G. Ellinas, “A data- driven bandwidth allocation framework with QoS considerations for EONs, ” IEEE/OSA J. Light. T echnol.37, 1853–1864 (2019)

  6. [6]

    Classification and forecasting of real- time server traffic flows employing long short-term memory for hybrid E/O data center networks,

    M. Balanici and S. Pachnicke, “Classification and forecasting of real- time server traffic flows employing long short-term memory for hybrid E/O data center networks, ” IEEE/OSA J. Opt. Commun. Netw.13, 85– 93 (2021)

  7. [7]

    Machine- learning-aided dynamic reconfiguration in optical DC/HPC networks,

    S. K. Singh, C.-Y . Liu, S. J. Ben Y oo, and R. Proietti, “Machine- learning-aided dynamic reconfiguration in optical DC/HPC networks, ” in Proc. IEEE Optical Network Design and Modelling (ONDM), (2022)

  8. [8]

    Federated learning for network traffic prediction,

    S. Behera, S. K. Panda, T . Panayiotou, and G. Ellinas, “Federated learning for network traffic prediction, ” in Proc. IFIP Networking Con- ference, (2024)

Show all 37 references
  1. [9]

    A fair fed- erated learning framework for collaborative network traffic prediction and resource allocation,

    S. K. Panda, T . Panayiotou, G. Ellinas, and S. Behera, “A fair fed- erated learning framework for collaborative network traffic prediction and resource allocation, ” in Proc. IEEE International Conference on Communications (ICC), (2025)

  2. [10]

    A deep learning method based on an attention mechanism for wireless network traffic predic- tion,

    M. Li, Y . Wang, Z. Wang, and H. Zheng, “A deep learning method based on an attention mechanism for wireless network traffic predic- tion, ” Ad Hoc Networks107, 102258 (2020)

  3. [11]

    In- traflow temporal correlation-based network traffic prediction,

    J. Lu, C. T ang, Z. Chen, J. Guo, A. Zou, W. Y ang, and C. T ang, “In- traflow temporal correlation-based network traffic prediction, ” Comput. Networks 256, 110913 (2025)

  4. [12]

    DenseNet-T ransformer: A deep learning method for spatialtemporal traffic prediction in optical fronthaul network,

    X. Qin, W. Zhu, Q. Hu, Z. Zhou, Y . Ding, X. Gao, and R. Gu, “DenseNet-T ransformer: A deep learning method for spatialtemporal traffic prediction in optical fronthaul network, ” Comput. Networks 253, 110674 (2024)

  5. [13]

    Live demonstration of autonomous link-capacity adjustment in optical metro-aggregation networks,

    M. Balanici, P . Safari, B. Shariati, A. Jafari, J. K. Fischer, and R. Fre- und, “Live demonstration of autonomous link-capacity adjustment in optical metro-aggregation networks, ” in Proc. IEEE/OPTICA Optical Fiber Communications Conference (OFC), (2024), pp. 1–3

  6. [14]

    Addressing traffic prediction uncertainty in multi-period planning optical networks,

    T . Panayiotou and G. Ellinas, “Addressing traffic prediction uncertainty in multi-period planning optical networks, ” in Proc. IEEE/OPTICA Op- tical Fiber Communications Conference (OFC), (2022)

  7. [15]

    Uncertainty quantifica- tion and consideration in ML-aided traffic-driven service provisioning,

    H. Maryam, T . Panayiotou, and G. Ellinas, “Uncertainty quantifica- tion and consideration in ML-aided traffic-driven service provisioning, ” Comput. Commun. 202, 13–22 (2023)

  8. [16]

    Quantifying the operational benefits of deep learning based dynamic traffic pre- diction using real-world dataset,

    D. Uzunidis, C. Christofodis, I. D. Francesca, J. M. R. Moscoso, D. Larrabeiti, J. M. Fabrega, D. M. Marom, and I. T omkos, “Quantifying the operational benefits of deep learning based dynamic traffic pre- diction using real-world dataset, ” in Proc. IEEE/OPTICA Optical Fiber Co...

  9. [17]

    Minimize sub- carrier reallocation in elastic optical path networks using traffic pre- diction,

    S. Shakya, Y . Wang, X. Cao, Z. Y e, and C. Qiao, “Minimize sub- carrier reallocation in elastic optical path networks using traffic pre- diction, ” inProc. IEEE Global Communications Conference (GLOBE- COM), (2013)

  10. [18]

    Multi-step traffic predic- tion for multi-period planning in optical networks,

    H. Maryam, T . Panayiotou, and G. Ellinas, “Multi-step traffic predic- tion for multi-period planning in optical networks, ” inProc. International Conference on T ransparent Optical Networks (ICTON), (2024)

  11. [19]

    Energy efficient dynamic optical routing for mobile metro-core networks under tidal traf- fic patterns,

    R. Alvizu, X. Zhao, G. Maier, Y . Xu, and A. Pattavina, “Energy efficient dynamic optical routing for mobile metro-core networks under tidal traf- fic patterns, ” IEEE/OSA J. Light. T echnol.35, 325–333 (2017)

  12. [20]

    Ma- chine learning assisted provisioning of time-varying traffic in translu- cent optical networks,

    A. Wodarczyk, A. Knapiska, P . Lechowicz, and K. Walkowiak, “Ma- chine learning assisted provisioning of time-varying traffic in translu- cent optical networks, ” IEEE Access12, 110193–110212 (2024)

  13. [21]

    Optimal and near-optimal alpha-fair re- source allocation algorithms based on traffic demand predictions for optical network planning,

    T . Panayiotou and G. Ellinas, “Optimal and near-optimal alpha-fair re- source allocation algorithms based on traffic demand predictions for optical network planning, ” IEEE/OSA J. Opt. Commun. Netw. 13, 53– 68 (2021)

  14. [22]

    Balancing efficiency and fairness in resource allocation for optical networks,

    T . Panayiotou and G. Ellinas, “Balancing efficiency and fairness in resource allocation for optical networks, ” IEEE T rans. Netw. Serv. Manag. 21, 389–401 (2024)

  15. [23]

    Fair resource allocation in optical net- works under tidal traffic,

    T . Panayiotou and G. Ellinas, “Fair resource allocation in optical net- works under tidal traffic, ” in Proc. IEEE Global Communications Con- ference (GLOBECOM), (2020)

  16. [24]

    Knowledge-based autonomous service provisioning in multi-domain elastic optical net- works,

    X. Chen, R. Proietti, H. Lu, A. Castro, and S. Y oo, “Knowledge-based autonomous service provisioning in multi-domain elastic optical net- works, ” IEEE Commun. Mag.56, 152–158 (2018)

  17. [25]

    Machine-learning-based prediction for re- source (re)allocation in optical data center networks,

    S. K. Singh and A. Jukan, “Machine-learning-based prediction for re- source (re)allocation in optical data center networks, ” IEEE/OSA J. Opt. Commun. Netw. 10, D12–D28 (2018)

  18. [26]

    A machine learning approach to mitigating fragmentation and crosstalk in space division multiplexing elastic optical networks,

    Y . Xiong, Y . Y ang, Y . Y e, and G. N. Rouskas, “A machine learning approach to mitigating fragmentation and crosstalk in space division multiplexing elastic optical networks, ” Opt. Fiber T echnol. 50, 99–107 (2019)

  19. [27]

    T raffic pre- diction in optical networks using graph convolutional generative ad- versarial networks,

    C. Vinchoff, N. Chung, T . Gordon, L. Lyford, and M. Aibin, “T raffic pre- diction in optical networks using graph convolutional generative ad- versarial networks, ” inProc. International Conference on T ransparent Optical Networks (ICTON), (2020)

  20. [28]

    T wo use cases of machine learning for SDN-enabled IP/optical networks: Traffic matrix prediction and optical path performance prediction,

    G. Choudhury, D. Lynch, G. Thakur, and S. Tse, “T wo use cases of machine learning for SDN-enabled IP/optical networks: Traffic matrix prediction and optical path performance prediction, ” IEEE/OSA J. Opt. Commun. Netw. 10, D52–D62 (2018)

  21. [29]

    Learning phrase representations using RNN encoder-decoder for statistical machine translation,

    K. Cho, B. Van Merriënboer, C. Gulcehre, D. Bahdanau, F . Bougares, H. Schwenk, and Y . Bengio, “Learning phrase representations using RNN encoder-decoder for statistical machine translation, ” arXiv:1406.1078 [cs.CL] (2014)

  22. [30]

    Time-series prediction with RNN, GRU, LSTM and attention,

    J. Y oon, “Time-series prediction with RNN, GRU, LSTM and attention, ” https://www.ericsson.com/en/reports-and-papers/mobility-report/ dataforecasts/mobile-traffic-forecast (2024). Accessed: 2026-02-10

  23. [31]

    Machine learning framework for timely soft-failure detection and localization in elastic optical net- works,

    S. Behera, T . Panayiotou, and G. Ellinas, “Machine learning framework for timely soft-failure detection and localization in elastic optical net- works, ” IEEE/OPTICA J. Opt. Commun. Netw. 15, E74–E85 (2023)

  24. [32]

    Finding the k shortest loopless paths in a network,

    J. Y . Y en, “Finding the k shortest loopless paths in a network, ” Manag. Sci. 17, 712–716 (1971)

  25. [33]

    Distance-adaptive spectrum resource allocation in spectrum-sliced elastic optical path network,

    M. Jinno, B. Kozicki, H. T akara, A. Watanabe, Y . Sone, T . T anaka, and A. Hirano, “Distance-adaptive spectrum resource allocation in spectrum-sliced elastic optical path network, ” IEEE Commun. Mag.48, 138–145 (2010)

  26. [34]

    Time-varying spectrum allocation policies and blocking analysis in flexible optical networks,

    K. Christodoulopoulos, I. T omkos, and E. Varvarigos, “Time-varying spectrum allocation policies and blocking analysis in flexible optical networks, ” IEEE J. Sel. Areas Commun. 31, 13–25 (2013)

  27. [35]

    High Performance Computing Facility of the University of Cyprus (UCY HPC),

    “High Performance Computing Facility of the University of Cyprus (UCY HPC), ” (2026). Computational resources provided by UCY HPC

  28. [36]

    Gurobi Optimizer Reference Manual,

    Gurobi Optimization, LLC, “Gurobi Optimizer Reference Manual, ” (2026)

  29. [37]

    Elastic band- width allocation in flexible OFDM-based optical networks,

    K. Christodoulopoulos, I. T omkos, and E.A. Varvarigos, “Elastic band- width allocation in flexible OFDM-based optical networks, ” IEEE/OSA Research Article 10 J. Light. T echn. 29, 1354–1366 (2011)

Pith tools

Reviewed June 29, 2026 · model on record in the stance chip above.