Experimenting with Energy-Awareness in Edge-Cloud Containerized Application Orchestration
Dalal Ali, Rute C. Sofia
TL;DR
This work tackles energy inefficiency in edge-cloud container orchestration by injecting cross-layer energy metrics into Kubernetes scheduling to achieve greener deployments. The authors introduce CODECO, a Kubernetes-compatible framework with PDLC-CA that aggregates network and compute energy signals into node-cost scores and provides energy-aware recommendations to the main SWM scheduler, using metrics such as $n_e(i)$, $l_e(i,j)$, and $L_e(i)$. Experimental evaluation on a real ARM-based testbed shows that CODECO can reduce total energy consumption and improve energy-efficiency, especially under high load, though energy-model limitations and overheads can affect gains under moderate load. These results suggest that cross-layer energy-awareness, along with flexible performance profiles, can yield more sustainable cloud-native deployments in heterogeneous edge-cloud environments.
Abstract
This paper explores the role of energy-awareness strategies into the deployment of applications across heterogeneous Edge-Cloud infrastructures. It proposes methods to inject into existing scheduling approaches energy metrics at a computational and network level, to optimize resource allocation and reduce energy consumption. The proposed approach is experimentally evaluated using a real-world testbed based on ARM devices, comparing energy consumption and workload distribution against standard Kubernetes scheduling. Results demonstrate consistent improvements in energy efficiency, particularly under high-load scenarios, highlighting the potential of incorporating energy-awareness into orchestration processes for more sustainable cloud-native computing.