Application and data placement solutions for workloads at the network edge

Sammanfattning

The growing proliferation of mobile and Internet of Things (IoT) devices has led to increasing demand for computational offloading to nearby infrastructure, making edge computing a critical paradigm for modern distributed systems. Computational offloading refers to transferring intensive processing tasks from resource-constrained devices to more powerful edge servers, aiming to improve application responsiveness and reduce device energy consumption. While offloading can reduce latency and improve service quality, placing applications and their associated data efficiently onto suitable edge nodes remains a complex challenge, especially in environments with user mobility and limited computational resources. In such cases, optimal placements can improve system performance and reduce the amount of data transferred across the network. This thesis investigates the problem of coordinating application and data placement in edge networks by developing migration strategies that adapt to changing conditions. The study explores the trade-off between minimizing network overhead and maintaining service responsiveness. To address this, a placement strategy based on Multi-Criteria Decision Making (MCDM) is proposed, which employs a genetic algorithm to optimize the weights of the different placement criteria. Numerical results show that the proposed algorithm consistently learns the optimal strategy, outperforming the baseline strategies. These findings highlight the importance of adaptive placement strategies that jointly consider both application placement and data migration in edge computing environments. By improving resource utilization and reducing communication overhead, the proposed method contribute to more efficient and responsive service delivery. This work lays the foundation for future research on adaptive placement mechanisms in edge network deployments.