Dynamic resource provisioning in containerized edge systems with reconfigurable edge servers

Recent technological advancements have seen powerful computational resource-enriched virtual machines (VMs) being used for processing data in edge servers. However, the high energy demands and excessive overhead associated with launching VMs are major obstacles to achieving energy-efficient operations in multi-access edge computing environments. As a result, there has been a relentless acceleration toward container virtualization to provide containerized services at the edge. The lightweight nature of containers compared to VMs makes them a popular technology for edge computing platforms. However, two significant challenges have been identified. The first is the problem of providing real-time support for containerized edge systems (to combat issues of high latency, anomaly detection, and automated monitoring and control, among others). The other problem is that, although containers help reduce application deployment time, considerable network bandwidth is expended and longer download queues are experienced on each node in the network. We propose a dynamic resource provisioning scheme for containerized edge systems to address these challenges. The proposed scheme employs containerized reconfigurable edge servers, which enable computational task operations to be moved to the data source for easier and quicker completion. Then, a novel adaptive power management technique based on predictive control through finite system observations is used to effectively estimate and regulate the energy consumption in the edge-based network. The adaptive controller schedules computational resources on a time slot basis in an adaptive manner, while continuing to receive updates to plan future resource provisioning. The proposed technique is evaluated using welfare gain, server response rate, and energy consumption metrics and is shown to outperform recent comparative models significantly.