Container-based Power Aware Resource Management in Cloud Systems

The rapid transition toward container-based cloud computing has intensified the demand for advanced resource management strategies that simultaneously enhance energy efficiency, ensure balanced load distribution, and maintain Service Level Agreement (SLA) compliance. Containers provide lightweight isolation and faster deployment compared to traditional hypervisor-based Virtual Machines (VMs), making them ideal for large-scale, dynamic cloud infrastructures. However, their adoption leads to higher migrations, greater SLA violation risk, and resource fragmentation. This paper proposes a Container-based Power-Aware Resource Management (CPARM) framework, an adaptive, multi-phase scheduling framework integrating CPU-aware initial container placement, non-linear real-time power modeling, overload detection, continuous SLA monitoring, and the Job-Co-location-Aware Migration algorithm. CPARM is implemented and evaluated in the ContainerCloudSim simulation environment using real-world PlanetLab workload traces. Experimental results demonstrate that the proposed CPARM framework outperforms existing algorithms, including FF, MF, LF, and EE-PSO, in terms of energy efficiency up to 58%, container migrations up to 88%, and SLA compliance up to 92%. This demonstrates CPARM potential as a highly effective solution for modern container-based cloud data centers.