Energy-Optimized Scheduling for AIoT Workloads Using TOPSIS

2025-06-05

AI Summary

To address the insufficient joint optimization of energy efficiency, latency, and resource balance for AIoT workloads in cloud–edge heterogeneous environments, this paper proposes GreenPod—the first energy-aware scheduling framework that integrates the TOPSIS multi-criteria decision-making method into the Kubernetes scheduler. Its key contributions include: (1) enabling dynamic weighted, five-dimensional joint optimization—comprising execution time, energy consumption, CPU/memory utilization, and load balancing—and (2) unifying sustainability with real-time performance. Extensive experiments on heterogeneous Google Kubernetes Engine (GKE) clusters demonstrate that GreenPod reduces energy consumption by up to 39.1% compared to the native Kubernetes scheduler, achieving optimal energy savings for medium-complexity AIoT tasks while incurring negligible increases in scheduling latency. These results validate GreenPod’s effectiveness and practicality in realistic edge-computing scenarios.

Technology Category

Machine Learning: Efficient ML / Green AIPlanning, Routing, and Scheduling: Optimization of Spatio-temporal SystemsConstraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environmentsEconomics, Online Markets and Human Computation: Fairness and ethical considerations in crowd work and in human-in-the-loop AI systemsResponsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

AIoT workloads demand energy-efficient orchestration across cloud-edge infrastructures, but Kubernetes' default scheduler lacks multi-criteria optimization for heterogeneous environments. This paper presents GreenPod, a TOPSIS-based scheduler optimizing pod placement based on execution time, energy consumption, processing core, memory availability, and resource balance. Tested on a heterogeneous Google Kubernetes cluster, GreenPod improves energy efficiency by up to 39.1% over the default Kubernetes (K8s) scheduler, particularly with energy-centric weighting schemes. Medium complexity workloads showed the highest energy savings, despite slight scheduling latency. GreenPod effectively balances sustainability and performance for AIoT applications.

Problem

• Optimizes energy-efficient AIoT workload scheduling in cloud-edge
• Addresses Kubernetes' lack of multi-criteria optimization for heterogeneous environments
• Balances sustainability and performance using TOPSIS-based pod placement

Innovation

• TOPSIS-based scheduler optimizes pod placement
• Considers energy, time, cores, memory, balance
• Improves energy efficiency by 39.1%