Multi-Factorial Energy Aware Resource Management in Edge Networks

Edge networks deliver computing services close to the user, unlike centralized clouds. This improves service scalability and delay-sensitive functions can be offloaded to the edge, when the latency incurred by cloud services is too high. Since services in edge networks, by their nature, are not centralized, careful design is required to achieve efficient resource utilization and low power consumption. These issues are addressed in this paper. A network device power model is formulated to explore the power dissipation characteristics of frequency scalable CMOS devices (as measured using a NetFPGA testbed). An on-demand energy-efficient resource allocation model (OERA) is designed based on this model. OERA features acceptance ratios that are 11%–17% higher than existing solutions and 9% lower power consumption. A novel algorithm is presented for resource placement in edge networks, which can accommodate higher traffic flow demands and distribution distance than existing solutions. This uses mixed integer linear programming to simultaneously maximize the aggregate flow demands and to minimize the network energy consumption. An iterative algorithm and a heuristic greedy edge network device placement algorithm are implemented that not only solve this NP-Hard problem but also significantly reduce the network energy consumption.