GREEN COMPUTING AND CARBON FOOTPRINT REDUCTION: A COMPREHENSIVE FRAMEWORK FOR SUSTAINABLE IT INFRASTRUCTURE MANAGEMENT

Abstract

Abstract— The explosive proliferation of digital services, cloud computing platforms, and artificial intelligence workloads has transformed the information and communication technology (ICT) sector into one of the most energy-intensive industries on the planet. Global data centers consumed an estimated 415 terawatt-hours of electricity in 2024, and projections indicate this figure will surpass 900 terawatt-hours by 2030 if present growth trajectories continue unchecked. Concurrently, the embodied carbon locked into semiconductor manufacturing—a dimension frequently neglected in sustainability reporting—can exceed the entire operational carbon of a computing device over its useful lifetime. This paper presents the Integrated Carbon Management Framework (ICMF), a five-layer architecture that delivers real-time visibility into all material sources of IT-related carbon emissions and converts that visibility into automated, policy-driven reduction actions. The framework encompasses operational electricity monitoring linked to live grid carbon intensity feeds, amortized embodied carbon tracking across full hardware lifecycles, software carbon intensity scoring aligned with the ISO/IEC 21031:2024 specification, and a carbon-aware workload scheduling algorithm that temporally and spatially shifts delay-tolerant jobs to minimize aggregate emissions without violating service-level agreements. Evaluation against six months of anonymized enterprise workload traces demonstrates aggregate operational carbon reductions of 21.8%, with temporal scheduling alone achieving 32.4% savings for batch workloads. Lifecycle-guided hardware replacement further reduces annual embodied carbon procurement by 18.3% relative to conventional warranty-based policies. A vendor-agnostic reporting protocol is additionally proposed to resolve the methodological divergence identified across leading cloud provider carbon calculators, reducing inter-calculator variance from 43% to below 5%. The paper concludes with a roadmap for extending these contributions to edge computing environments and federated sustainability benchmarking.