Learning When to Wait: A Lightweight Binary Decision Framework for Container Migration in Edge Computing

Container migration is one of the important techniques used to provide the continuity of services in Multi-access Edge Computing environments. Services are migrated from edge servers to follow the mobility of users among the available edge servers. Nevertheless, current strategies do not take into account the cost that a container migration causes in terms of downtime, bandwidth consumption, energy cost, and the possible occurrence of failures during the process. In this work, a lightweight pre-migration binary decision framework is introduced that works as a gatekeeper before any migration attempt. According to the actual state of the system, the framework decides at each epoch whether to migrate or not, classifying the epochs into Migrate and Wait categories. The training labels are obtained through a counterfactual approach using simulated data generated using the EdgeSimPy simulation tool. The Decision Tree classifier has been chosen as the main classifier due to its minimal computing complexity and interpretability. The effectiveness of the suggested framework is measured on the basis of comparison with three baseline policies, including always migrate, never migrate, and a latency-threshold rule, along with multiple machine learning models and deep learning models, and such metrics as accuracy, precision, recall, F1-score, AUC, and runtime performance metrics. It is shown that the suggested solution outperformed all three baseline policies along with the machine learning and deep learning models, proving that an intelligent decision-making stage is essential in Multi-access Edge Computing.

@article{Abu-Haltem2026,
  title     = {Learning When to Wait: A Lightweight Binary Decision Framework for Container Migration in Edge Computing},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {17},
  number    = {7},
  year      = {2026},
  publisher = {The Science and Information Organization},
  author    = {Yasmeen Abu-Haltem and Sawsan AlShattnawi},
  doi       = {10.14569/IJACSA.2026.0170786},
  url       = {https://doi.org/10.14569/IJACSA.2026.0170786}
}