Predictive and Power-Aware Application Migration in MEC-Enabled 5G Vehicular Networks

The quality of service for delay-sensitive applications in vehicular networks can be greatly improved with the adoption of 5G networks and Multi-access Edge Computing (MEC). MEC allows vehicular applications to be hosted on virtualized infrastructure at the edge of the network in close proximity to the user rather than in remote cloud data centers. This provides improved computing power compared to onboard vehicular resources, while keeps end-to-end latency low. However, the highly dynamic nature of vehicular networks, frequent handovers between base stations, and the resource constraints of MEC hosts pose significant challenges in maintaining service continuity. To handle these challenges, this thesis first proposes a proactive migration strategy that leverages Signal-to-Interference-plus- Noise Ratio (SINR) values and employs a sequence-to-sequence Long Short-Term Memory (LSTM) model to predict handovers proactively and trigger timely migration, aiming to reduce end-to-end delay during the migration process time. Following this, it introduces a congestion window adaptation mechanism, based on the MEC Radio Network Information Service (RNIS) data, to enhance Transmission Control Protocol (TCP) performance after migration. In addition to improving service continuity, this research proposes an Energy-Efficient Proactive Migration Algorithm (EEPA) that integrates power-awareness into MEC application migration orchestration in order to optimize energy consumption while maintaining high performance. EEPAdynamically evaluates the CPU load, power consumption, and resource availability of predicted target hosts to make migration decisions. This approach considers an application awareness strategy that enables it to differentiate between delay-sensitive and delay-tolerant applications. Using this, the proposed solution prioritizes latency-critical services’ migration and maintains the overall energy efficiency. System-level simulations demonstrate that each proposed component improves specific aspects of system performance: the predictive migration mechanism significantly reduces end-to-end delay compared to traditional reactive approaches, while the congestion control strategy prevents unnecessary TCP throughput degradation across diverse network conditions. Furthermore, the energy-efficient migration algorithm effectively balances power consumption with service continuity, achieving up to 6% power savings without compromising latency performance compared to other algorithms. Additionally, it ensures the network is able to maintain round-trip time low for delay-sensitive applications. In conclusion, these results highlight the framework’s capability to ensure seamless service for delay-sensitive vehicular applications, such as real-time video streaming, while optimizing resource utilization at MEC hosts.