Preemptive SDN Load Balancing with Machine Learning for Delay Sensitive Applications

TitrePreemptive SDN Load Balancing with Machine Learning for Delay Sensitive Applications
Publication TypeJournal Article
Year of Publication2020
AuthorsFilali, A, Mlika, Z, Cherkaoui, S, Kobbane, A
JournalIEEE Transactions on Vehicular Technology
Mots-clés5G mobile communication systems, Auto-regressive integrated moving average, Autoregressive moving average model, Balancing, Comparative studies, Controllers, Delay constraints, Delay-sensitive applications, Forecasting, Learning algorithms, Long short-term memory, Long-term prediction, Low-latency communication, Prediction errors, Predictive analytics, Reactive mechanism, Reinforcement learning

SDN is a key-enabler to achieve scalability in 5G and Multi-access Edge Computing networks. To balance the load between distributed SDN controllers, the migration of the data plane components has been proposed. Different from most previous works which use reactive mechanisms, we propose to preemptively balance the load in the SDN control plane to support network flows that require low latency communications. First, we forecast the load of SDN controllers to prevent load imbalances and schedule data plane migrations in advance. Second, we optimize the migration operations to achieve better load balancing under delay constraints. Specifically, in the first step, we construct two prediction models based on Auto Regressive Integrated Moving Average (ARIMA) and Long Short-Term Memory (LSTM) approaches to forecast SDN controllers load. Then, we conduct a comparative study between these two models and calculate their accuracies and forecast errors. The results show that, in long-term predictions, the accuracy of LSTM model outperforms that of ARIMA by 55% in terms of prediction errors. In the second step, to select which data plane components to migrate and where the migration should happen under delay constraints, we formulate the problem as a non-linear binary program, prove its NP-completeness and propose a reinforcement learning algorithm to solve it. The simulations show that the proposed algorithm performs close to optimal and outperforms recent benchmark algorithms from the literature. © 1967-2012 IEEE.




Suivez-nous sur





Avenue Mohammed Ben Abdallah Regragui, Madinat Al Irfane, BP 713, Agdal Rabat, Maroc

  Télécopie : (+212) 5 37 68 60 78

  Secrétariat de direction : 06 61 48 10 97

        Secrétariat général : 06 61 34 09 27

        Service des affaires financières : 06 61 44 76 79

        Service des affaires estudiantines : 06 62 77 10 17 /

        Résidences : 06 61 82 89 77



    Compteur de visiteurs:483,470
    Education - This is a contributing Drupal Theme
    Design by WeebPal.