Pith. sign in

REVIEW 5 major objections 5 minor 22 references

Optimizing Server Load Distribution in Multimedia IoT Environments through LSTM-Based Predictive Algorithms

T0 review · 5 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Predictive load balancing using LSTM forecasts and fuzzy classification keeps servers below overload and saves energy in software-defined multimedia IoT networks.

desk verdict A coherent but thin incremental extension of the authors' own LSTM/fuzzy load-balancing work; the evaluation is too confounded and internally inconsistent to support the advertised energy and QoS gains. read the letter →

arxiv 2505.24806 v1 pith:QM4PFWLM submitted 2025-05-30 cs.NI

Somaye Imanpour , Ahmadreza Montazerolghaem , Saeed Afshari This is my paper
classification cs.NI
keywords InternetofMultimediaThingsSoftware-DefinedNetworkingserverloadbalancingLSTMpredictionfuzzyclassificationenergyefficiencyoverloadpreventionresourceutilizationforecasting
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 aims to show that a software-defined multimedia IoT network can keep its servers below overload by acting on forecasts rather than current readings. It proposes measuring each server's CPU, memory, hard disk, and bandwidth use as normalized values, feeding those time series to an LSTM network, and then sorting servers into four fuzzy load levels: under-load, normal-load, highly-load, and over-load. New traffic is sent first to under-loaded servers, and overloaded or highly-loaded servers shed work to servers that can absorb it without moving into overload. In a Mininet and Floodlight simulation with 26 growing UDP flows, the authors report that no server reaches over-load under the proposed scheme, while random and round-robin selection push several servers into over-load. If correct, the method would improve service quality and cut energy costs by powering off under-used servers.

What carries the argument

The load-bearing mechanism is the pairing of an LSTM time-series forecast with a Mamdani fuzzy classifier whose output drives a greedy migration algorithm. LSTM is a recurrent neural network that learns temporal dependencies from historical resource readings; here it is trained with Adam, MSE loss, 15 epochs, and batch size one, with input normalized by MinMaxScaler. The fuzzy classifier converts the four predicted ratios into linguistic values (low, medium, high) through membership functions and applies 81 rules to output one of four load levels. This predicted category, not the current measured load, is what selects target servers for new flows and for load migration, and the same classification decides when to shut a server down.

What would settle it

Re-run the 26-flow scenario while recording both the LSTM-predicted resource values and the measured values at each timestep, then count how often the fuzzy load level assigned from predictions differs from the level assigned from measurements; if misclassifications push a server above the over-load boundary, the method's safety depends on forgiving thresholds rather than on accurate forecasts.

Watch

Extended reading notes

Core claim

The central claim is that predictive load balancing based on LSTM forecasts plus a four-level Mamdani fuzzy classifier keeps servers in software-defined multimedia IoT networks below the over-load threshold while saving energy. Server load is defined by four normalized resource ratios: $x_i$ for CPU, $y_i$ for memory, $z_i$ for hard disk, and $w_i$ for bandwidth, each current usage divided by capacity as in Equations (1)-(4). The controller forms a $4 \times W$ consumption matrix $f_{h_i}$ for each server and feeds it to a single-layer LSTM with five neurons; lookback is five for CPU, disk, and bandwidth and two for memory. The predicted values $\hat{x}_{t+1}^i$, $\hat{y}_{t+1}^i$, $\hat{z}_{t+1}^i$, $\hat{w}_{t+1}^i$ are mapped through 81 if-then fuzzy rules to one of four levels, and Algorithm 1 routes flows to under-load servers, transfers load away from over-load and highly-load servers, and powers off under-load servers. The reported simulation claims that all four servers stay below over-load across 26 flows, whereas random and round-robin selection leave at least two servers at over-load in the same scenario.

Load-bearing premise

The load-bearing premise is that the LSTM's predicted CPU, memory, disk, and bandwidth usage is accurate enough to put each server in the right fuzzy load class, even though the paper's own memory forecast on server 2 has $R^2=0.51$ and would be used without any uncertainty margin.

Editorial extensions

If this is right

  • An SDN controller can move traffic before a server actually overloads, because the LSTM forecast gives a lookahead window of one timestep.
  • Under-used servers can be switched off, and their switches with them, cutting energy without sacrificing headroom.
  • Because load is judged on four resources, a server with spare CPU but exhausted memory is not mistakenly treated as lightly loaded.
  • The four-level classification lets the controller apply graduated responses: moderate action for highly-loaded servers and full migration only for over-loaded ones.
  • The scheme remains stable under steadily increasing multimedia traffic: in the reported scenario all servers end at highly-load rather than over-load.

Reading between the lines

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

  • The paper leaves open whether forecast error near a fuzzy boundary could send traffic to a server that is actually closer to overload than predicted; propagating the reported per-metric errors into the classification would quantify this risk.
  • Because the baselines are random and round-robin selection, the experiments do not isolate the forecasting contribution; comparing the same fuzzy policy on measured current loads against the LSTM-based policy would do so.
  • A testable extension is to feed the fuzzy system prediction intervals instead of point forecasts, hedging classification decisions when the LSTM is uncertain, particularly for memory usage, the metric with the weakest reported fit.
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

5 major / 5 minor

Summary. The paper proposes a server load-balancing scheme for software-defined multimedia IoT networks. The controller measures CPU, memory, disk, and bandwidth utilization of each server, uses LSTM to forecast the next-step values of these four metrics, and then applies a four-level fuzzy classifier (under-load, normal-load, highly-load, over-load) to decide where to route new flows and whether to migrate traffic. The algorithm also powers off under-loaded servers and their associated switches to save energy. The evaluation is performed in a Mininet/Floodlight emulation with four servers of different capacities and 26 UDP flows of increasing size; the proposed method is compared with random and round-robin server selection, and the results are shown as qualitative load-level trajectories. The paper also reports LSTM prediction quality (RMSE, MAE, R2) for the four metrics on server 2 and concludes that the approach improves load distribution, service quality, and energy efficiency.

Significance. If validated, the proposed architecture could be a useful proactive load-balancing mechanism for SDN-based multimedia IoT networks, and the use of four resource metrics with a four-level fuzzy classifier is a reasonable extension of earlier three-level schemes. Credit is due for the explicit Algorithm 1, the clear testbed description, and the effort to apply LSTM to server load prediction. However, the current evidence does not support the central claim: the predictive component is not isolated from the power-management logic, no quantitative QoS or energy metrics are reported, and the reported prediction quality is modest. As it stands, the paper is better viewed as a system sketch with a preliminary demonstration rather than a validated contribution.

major comments (5)
  1. [Section 4, Figures 11-13] The comparison does not isolate the predictive component. The proposed method includes explicit power management: under-loaded servers and their switches are turned off, as described in the text around Figure 11. The random and round-robin baselines are described only as selection schemes, with no mention of equivalent power-down or migration logic. Therefore, the qualitative advantage shown in Figures 11-13 (no overload versus overload) could be entirely due to the power-management or migration policy, not to LSTM forecasting. Furthermore, the paper reports no quantitative metrics for QoS, throughput, latency, or energy consumption, so the abstract's claims of 'improved service quality, reduced operational costs, and increased productivity' are not directly measured. The authors should rerun the evaluation with identical power-management and migration logic across all methods, include an ablation without prediction (e.g., using current loads) and with a simpler predictor, and report numerical metrics such as overload count, energy consumption, and a QoS measure.
  2. [Sections 3.1 and 4] The LSTM architecture is specified twice with contradictory parameters. Section 3.1 (under 'Model Architecture Details') states a single LSTM layer with five neurons, lookback 5 (or 2 for memory), batch size 1, and 15 epochs. Section 4 states 'The LSTM model consists of two hidden layers, each with 64 hidden units. A batch size of 32 was used during training.' These configurations cannot both describe the experiments that produced Figures 14-17, and the paper does not indicate which configuration was used. This contradiction makes the evaluation unreproducible and must be resolved in a revision.
  3. [Section 4, dataset description] The dataset description is a placeholder: 'The dataset consists of approximately N records spanning a time period of T hours/days.' This gives no basis for judging the statistical validity of the prediction results. The authors must provide the actual number of records, the time span, the collection environment, and the exact train/test split. They should also report prediction errors for all four servers and all four metrics, not only server 2, because the load-balancing decisions in Algorithm 1 depend on forecasts for every server.
  4. [Section 1 and Section 4] The claim that LSTM 'has been shown to outperform traditional methods such as NLMS and ANN' is supported only by citation [4], which is a survey of deep neural networks for stock market prediction, not for server load balancing. The evaluation in this paper does not compare LSTM against NLMS or ANN on the server-load data. In addition, the comparison in Section 4 excludes the authors' own earlier NLMS-based method [8] and the earlier conference version of this work [19], which are the most relevant baselines. The authors should either include these baselines in the evaluation or substantially soften the superiority claim.
  5. [Section 3, 'Determining Server Load Levels Using Fuzzy Systems' and Section 4] The fuzzy classifier takes the predicted values directly as inputs without any treatment of forecast uncertainty. The reported prediction quality for memory on server 2 is R2 = 0.51, which the paper itself describes as 'good' but which leaves a large portion of variance unexplained. Since the load-level classification is entirely determined by these forecasts, the paper should report the classification accuracy of the four-level fuzzy system against ground-truth levels (e.g., a confusion matrix) and discuss how prediction errors propagate to misclassification and suboptimal balancing decisions.
minor comments (5)
  1. [Section 4, Figures 12 and 13] The captions are mislabeled: Figure 12 is described in the text as the random selection scheme but its caption reads 'Server Load Levels in the Round-Robin Selection Scheme,' and Figure 13 has the same round-robin caption though the text says it shows round-robin. Please correct the captions and the numbering.
  2. [Section 1, Introduction] The first two paragraphs of the introduction are duplicated verbatim; remove the duplicate passage.
  3. [Table 2] The capacity symbols are swapped in the table: 𝜗𝑖^𝑧 is described as 'Bandwidth resource capacity' and 𝜗𝑖^𝑤 as 'Hard disk resource capacity,' whereas Equations (3) and (4) use z_i for hard disk and w_i for bandwidth. The same swap appears in the descriptions of f_hdd and f_bw.
  4. [Section 3.1, Equation (5)] The matrix in Equation (5) contains a malformed entry ('𝑥𝑃1 𝑥𝑃1 …') and the condition '𝑖𝑓 𝑀́ < 𝑃' uses an undefined variable Ḿ; please clarify the notation and the intended structure of the matrix.
  5. [General] The paper does not state how this journal submission extends the authors' conference paper [19]; briefly describing the difference would help situate the contribution.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the LSTM/fuzzy balancing chain is not derived from its own outputs, though the comparison suffers from missing baseline controls.

full rationale

The derivation chain is self-contained in the sense required for circularity analysis. Equations (1)-(4) normalize measured resource consumption; Equation (5) assembles those values into a matrix; the LSTM forecasts x, y, z, and w on a held-out test set with reported RMSE/MAE/R2; the fuzzy system's 81 rules map those forecasts to four load classes; Algorithm 1 then selects destination servers by those classes and re-evaluates post-migration levels. No fitted parameter is renamed as a prediction, no load-balancing outcome is used to define the LSTM inputs, and no uniqueness theorem is imported from the authors' earlier work to force the choice of method. The self-citations ([8], [13], [15], [19]) are used as related-work contrasts or as labels for the Mininet/Floodlight tools (with apparent reference-numbering errors), not as the load-bearing justification for prediction accuracy or the load-balancing effect. The main weaknesses are evidentiary rather than circular: random and round-robin baselines are not given the same power-management/migration logic, no QoS or energy metrics are reported, the dataset description is a placeholder ('approximately N records spanning a time period of T hours/days'), and the claimed LSTM superiority over NLMS/ANN is imported from a stock-market survey [4] rather than measured here. These are correctness, reproducibility, and support concerns, not cases where the derivation reduces to its own inputs.

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

The central claim rests on hand-chosen LSTM and fuzzy parameters, plus assumptions that measurements and migrations are effectively free and that forecasts are accurate. No new physical or computational entity is postulated.

free parameters (3)
  • LSTM lookback window = 5 for CPU/disk/bandwidth, 2 for memory
    Selected by trial-and-error (Section 4); directly affects the prediction inputs and reported R2 values.
  • LSTM hidden configuration and training hyperparameters = Contradictory: 1 layer x 5 units, 15 epochs, batch 1 versus 2 layers x 64 units, batch 32, learning rate 0.001
    Both descriptions appear in Section 4; no single reproducible configuration is given.
  • Fuzzy membership boundaries and output range = Not fully specified; output class range [-0.6, 0.4]
    The 81 fuzzy rules and the membership functions in Figures 3-7 are hand-defined, and no sensitivity analysis is provided.
assumptions (3)
  • domain assumption The SDN controller can accurately read CPU, memory, disk, and bandwidth usage from every server and can migrate traffic with negligible cost or delay.
    Invoked in Section 3.1 and Algorithm 1; migration overhead and measurement error are not modeled.
  • domain assumption Normalized resource forecasts in [0,1] are accurate enough that classifying a server from predicted values reliably reflects its future load.
    Used in the fuzzy classification step; the paper reports R2 as low as 0.51 for memory but does not propagate prediction error into the load-balancing results.
  • domain assumption Iperf UDP flows of 10-260 MB reproduce the traffic properties of multimedia IoT workloads.
    Evaluation scenario in Section 4; no real IoMT trace, packet-level QoS, or application timing is used.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Optimizing Server Load Distribution in Multimedia IoT Environments through LSTM-Based Predictive Algorithms." pith.science (2026). https://pith.science/paper/QM4PFWLM

@misc{pith2026250524806,
  author       = {Pith},
  title        = {Pith review of: Optimizing Server Load Distribution in Multimedia IoT Environments through LSTM-Based Predictive Algorithms},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QM4PFWLM}},
  note         = {Machine review of arXiv:2505.24806}
}
read the original abstract

The Internet of Multimedia Things (IoMT) represents a significant advancement in the evolution of IoT technologies, focusing on the transmission and management of multimedia streams. As the volume of data continues to surge and the number of connected devices grows exponentially, internet traffic has reached unprecedented levels, resulting in challenges such as server overloads and deteriorating service quality. Traditional computer network architectures were not designed to accommodate this rapid increase in demand, leading to the necessity for innovative solutions. In response, Software-Defined Networks (SDNs) have emerged as a promising framework, offering enhanced management capabilities by decoupling the control layer from the data layer. This study explores the load balancing of servers within software-defined multimedia IoT networks. The Long Short-Term Memory (LSTM) prediction algorithm is employed to accurately estimate server loads and fuzzy systems are integrated to optimize load distribution across servers. The findings from the simulations indicate that the proposed approach enhances the optimization and management of IoT networks, resulting in improved service quality, reduced operational costs, and increased productivity.

Figures

Figures reproduced from arXiv: 2505.24806 by the authors.

Figure 1
Figure 1. Software-Defined Multimedia Internet of things Network Model As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Flowchart of the Proposed Method Algorithm 3.1. System Model The network comprises switches, servers, controllers, and links. Let H={h1,h2,…,hi,…,hM } represent the set of servers, where M indicates the total number of servers in the network, and hi denotes the ith server. For clarity, [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Membership Function Input for 𝑥̂𝑡+1 𝑖 in the Range [0,1]} [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (14 more)
Figure 4
Figure 4. Figure 4: Membership Function Input for 𝑦̂𝑡+1 𝑖 in the Range [0,1]} [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Membership Function Input for 𝑧̂𝑡+1 𝑖 in the Range [0,1]} [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Membership Function Input for 𝑤̂𝑡+1 𝑖 in the Range [0,1]} • Fuzzy Rules Definition: The fuzzy rules necessary for classifying servers are defined. These rules encompass conditions that depend on the features and their combinations. The system comprises 81 rules defined…
Figure 7
Figure 7. Figure 7: Output Membership Function for Class Range [-0.6, 0.4]} • Final Classification: By combining various fuzzy outputs and employing post-processing logic, servers are categorized into different groups. The fuzzy system classifies servers into four levels: under-load, norm…
Figure 8
Figure 8. Figure 8: Network Topology for Performance Evaluation of the Server Load Balancing Algorithm In this section, as illustrated in [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: flow size Distribution in the Server Load Balancing Scenario. In this section, to simulate multimedia flow in software-defined IoT networks, the Iperf tool was used to establish a UDP connection from the client to the other servers. Using this method, multimedia flows …
Figure 10
Figure 10. Figure 10: Number of Datagrams Sent per flow. In this network, the servers have different capacities [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]
Figure 11
Figure 11. Figure 11: Server Levels in the Load Balancing Scenario At this stage, the proposed scheme is compared with two other methods for server selection: random selection and round-robin selection. These methods were chosen due to their widespread use as baseline techniques in load ba…
Figure 12
Figure 12. Figure 12: Server Load Levels in the Round-Robin Selection Scheme [PITH_FULL_IMAGE:figures/full_fig_p016_12.png]
Figure 13
Figure 13. Figure 13: also shows the server levels after executing the round-robin selection scheme. Server four reaches the over-load level after processing flow 13; however, despite this status, three other streams are sent to this server. In the final stream, which is flow 26, both the …
Figure 14
Figure 14. Figure 14: Predicted CPU Usage of the Second Server Using the LSTM Algorithm In [PITH_FULL_IMAGE:figures/full_fig_p017_14.png]
Figure 15
Figure 15. Figure 15: Predicted Memory Usage of the Second Server Using the LSTM Algorithm [PITH_FULL_IMAGE:figures/full_fig_p017_15.png]
Figure 16
Figure 16. Figure 16: Predicted Hard Disk Usage of the Second Server Using the LSTM Algorithm In [PITH_FULL_IMAGE:figures/full_fig_p018_16.png]
Figure 17
Figure 17. Figure 17: Predicted Bandwidth Usage of the Second Server Using the LSTM Algorithm 5. Conclusions The focus of this paper was to establish effective load distribution among servers within software-driven multimedia Internet￾of-Things (IoT) networks. The suggested method successf…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages

  1. [19]

    (2024, April)

    Imanpour, S., Montazerolghaem, A., & Afshari, S. (2024, April). Load balancing of servers in software -defined internet of multimedia things using the long short-term memory prediction algorithm. In 2024 10th International Conference on web research (ICWR) (pp. 291-296). IEEE

    work page 2024

  2. [8]

    Software-defined Internet of Multimedia Things: Energy -efficient and Load-balanced Resource Management

    A.Montazerolghaem , “Software-defined Internet of Multimedia Things: Energy -efficient and Load-balanced Resource Management.” IEEE Internet of Things Journal, vol. 9, 2021, 2432-2442

    work page 2021

  3. [4]

    A comprehensive survey on deep neural networks for stock market: The need, challenges, and futu re directions,

    A. Thakkar and K. Chaudhari, "A comprehensive survey on deep neural networks for stock market: The need, challenges, and futu re directions," Expert Systems with Applications, vol. 177, p. 114800, 2021

    work page 2021

  4. [1]

    Multimedia Internet of Things: A comprehensive surve y

    A.Nauman, Y. A.Qadri, M.Amjad , Y.B. Zikria, M.K.Afzal, and S.W. Kim, . “Multimedia Internet of Things: A comprehensive surve y.” Ieee Access, vol.8, 2020, pp.8202-8250

    work page 2020

  5. [2]

    Internet of multimedia things: Vision and challenges

    S. A. Alvi, B.Afzal, G. A.Shah, L.Atzori, and W.Mahmood, “Internet of multimedia things: Vision and challenges.” Ad Hoc Networks, vol.33, 2015, pp. 87-111

    work page 2015

  6. [3]

    SDN Controllers: Benchmarking & Performance Evaluation

    L. Zhu, M. M. Karim, K. Sharif, F. Li, X. Du, and M. Guizani, "SDN controllers: Benchmarking & performance evaluation," arXiv preprint arXiv:1902.04491, 2019

    work page Pith review arXiv 1902

  7. [5]

    Fuzzy sets,

    L. A. Zadeh, "Fuzzy sets," Information and control, vol. 8, no. 3, 1965 pp. 338-353

    work page 1965

  8. [6]

    Round -robin based load balancing in Software Defined Networking,

    S. Kaur, K. Kumar, J. Singh, and N. S. Ghumman, "Round -robin based load balancing in Software Defined Networking," in 2015 2nd international conference on computing for sustainable global development (INDIACom), 2015: IEEE, pp. 2136-2139

    work page 2015

Show all 22 references
  1. [7]

    LBBSRT: An efficient SDN load balancing scheme based on server response time,

    H. Zhong, Y. Fang, and J. Cui, "LBBSRT: An efficient SDN load balancing scheme based on server response time," Future Generation Computer Systems, vol. 68, 2017, pp. 183-190

    2017

  2. [9]

    Load balancing for network resource management i n software-defined networks,

    P. Abhishek, A. Naik, P. Doddannavar, R. Patil, M. M. Raikar, and S. Meena, "Load balancing for network resource management i n software-defined networks," in Advances in Distributed Computing and Machine Learning: Proceedings of ICADCML 2022: Springer, 2022, pp. 193-203

    2022

  3. [10]

    Proactive load balancing strategy towards intelligence -enabled software-defined network,

    C. Fancy and M. Pushpalatha, "Proactive load balancing strategy towards intelligence -enabled software-defined network," Arabian Journal for Science and Engineering, 2021, pp. 1-8

    2021

  4. [11]

    Adaptive Server Load Balancing in SDN Using PID Neural Network Controller,

    R. Malavika and M. L. Valarmathi, "Adaptive Server Load Balancing in SDN Using PID Neural Network Controller," Computer Syste ms Science & Engineering, vol. 42, no. 1, 2022

    2022

  5. [12]

    SDN-based hybrid server and link load balancing in multipath distributed storage systems,

    L. Guillen, S. Izumi, T. Abe, T. Suganuma, and H. Muraoka, "SDN-based hybrid server and link load balancing in multipath distributed storage systems," in NOMS 2018-2018 IEEE/IFIP Network Operations and Management Symposium, 2018: IEEE, pp. 1-6

    2018

  6. [13]

    Efficient resource allocation for multimedia streaming in software-defined internet of vehicles,

    A. Montazerolghaem, "Efficient resource allocation for multimedia streaming in software-defined internet of vehicles," IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 12, 2023, pp. 14718-14731

    2023

  7. [14]

    Optimized software -defined multimedia framework: networking and computing resource management,

    A. Montazerolghaem, "Optimized software -defined multimedia framework: networking and computing resource management," Journal of Ambient Intelligence and Humanized Computing, vol. 14, no. 9, 2023, pp. 12981-13001

    2023

  8. [15]

    Green cloud multimedia networking: NFV/SDN based energy-efficient resource allocation,

    A. Montazerolghaem, M. H. Yaghmaee, and A. Leon-Garcia, "Green cloud multimedia networking: NFV/SDN based energy-efficient resource allocation," IEEE Transactions on Green Communications and Networking, vol. 4, no. 3, 2020, pp. 873-889

    2020

  9. [16]

    Hybrid SDN networks: A multi -parameter server load balancing scheme,

    T. Malbašić, P. D. Bojović, Ž. Bojović, J. Šuh, and D. Vujošević, "Hybrid SDN networks: A multi -parameter server load balancing scheme," Journal of Network and Systems Management, vol. 30, no. 2, 2022, p. 30

    2022

  10. [17]

    Software defined networks (SDN): experimentation with Mininet topologies,

    D. Kumar and M. Sood, "Software defined networks (SDN): experimentation with Mininet topologies," Indian Journal of Science and Technology, vol. 9, no. 32, 2016, pp. 1-7

    2016

  11. [18]

    Experimenting with scalability of floodlight controller in software defined networks,

    S. Asadollahi and B. Goswami, "Experimenting with scalability of floodlight controller in software defined networks," in 2017 International Conference on Electrical, Electronics, Communication, Computer, and Optimization Techniques (ICEECCOT), 2017: IEEE, pp. 288-292

    2017

  12. [20]

    Task scheduling and load balancing in SDN -based cloud computing: A review of relevant research

    Mahdizadeh, Masoumeh, Ahmadreza Montazerolghaem, and Kamal Jamshidi. "Task scheduling and load balancing in SDN -based cloud computing: A review of relevant research." Journal of Engineering Research (2024)

    2024

  13. [21]

    Salehnia, T., MiarNaeimi, F., Izadi, S., Ahmadi, M., Montazerolghaem, A., Mirjalili, S., & Abualigah, L. (2024). A MTIS metho d using a combined of whale and moth-flame optimization algorithms. In Handbook of Whale Optimization Algorithm (pp. 625-651). Academic Press

    2024

  14. [22]

    R., & Yaghmaee, M

    Montazerolghaem, A. R., & Yaghmaee, M. H. (2013). SIP overload control testbed: Design, building and Evaluation. arXiv preprint arXiv:1307.3411

    2013 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.