Authors:
J. Biju, R. Senthilkumar, R. S. Kamalakannan, J. Senthil, Asique Kunhalakath

Addresses:
Division of Data Science and Cyber Security, Karunya Institute of Technology and Sciences, Coimbatore, Tamil Nadu, India. Department of Artificial Intelligence and Data Science, Shree Venkateshwara Hi-Tech Engineering College, Erode, Tamil Nadu, India. Department of Electronics and Communication Engineering, Shree Venkateshwara Hi-Tech Engineering College, Erode, Tamil Nadu, India. Department of Artificial Intelligence and Data Science, Center for Quantum Computing, Kalaignarkarunanidhi Institute of Technology (KIT), Coimbatore, Tamil Nadu, India. Department of Information Technology and Digital Transformation, The Zubair Corporation LLC, Muscat, Muscat Governorate, Oman.

Abstract:

The goal of this work is to develop a new orchestration framework aimed at reducing the carbon footprints of large-scale clouds while adhering to service level agreements SLAs and without increasing operational costs. Because of the growing share of electricity consumed by data centers worldwide, static workload scheduling approaches cannot fully account for temporal and spatial variations in the grid's carbon intensity. Our work proposes a dynamic scheduling algorithm for batch workloads subject to carbon footprint variability that pushes computing tasks to regions with lower carbon emissions, while meeting stringent latency requirements for user-centric applications. The experiments use a dataset of 484 jobs from a modified Google Cluster Trace, scheduled with Kubernetes for container scheduling and with Prometheus for real-time metric scraping. Researchers adopt a twin-objective optimization model in which spot instance cost is minimized while real-time carbon intensity is constrained. Results show that our approach significantly outperforms baseline schedulers in terms of carbon emissions, with approximately the same cost-effectiveness and performance stability. This paper reaffirms the belief that both environmental sustainability and economic viability are attainable in hyperscale cloud computing through intelligent, data-driven orchestration.

Keywords: Green Computing; Cloud Orchestration; Carbon Intensity; Cost Optimization; Kubernetes Scheduling; Dynamic Scheduling Algorithm; Service Level Agreements; Environmental Sustainability; Economic Viability.

Received on: 09/04/2025, Revised on: 06/06/2025, Accepted on: 11/09/2025, Published on: 12/06/2026

DOI: 10.69888/FTSCS.2026.000685

FMDB Transactions on Sustainable Computing Systems, 2026 Vol. 4 No. 2, Pages: 97-105

• Views : 49
• Downloads : 8

Download PDF