Green HPC: Carbon-Aware Scheduling in Cloud Data Centers

Authors:

Abstract
High-Performance Computing (HPC) has become a major force in scientific research, financial modeling, artificial intelligence and large-scale data analytics. Nevertheless, the growth of cloud-based HPC has led to severe environmental issues, particularly in terms of energy use and carbon emissions from hyperscale data centres. The paper explores HPC scheduling approaches to carbon consciousness under cloud configurations; attention is on the minimization of the Greenhouse Gas (GHG) emissions and achieving the associated systems performance and Service-Level Agreements (SLAs). It reviews recent approaches in green computing, discusses issues of incorporating renewable energy sources as they relate to scheduling policies and outlines a research approach that integrates workload prediction, carbon-intensity forecasting and multi-objective optimization. Simulation results indicate that this will reduce carbon emissions by 20-40 percent with minimal effect on the time of job completion. We examine the tradeoffs between energy efficiency and performance, as well as carbon consciousness, in the context of the rising prominence of sustainability metrics in the management of cloud data centres

This publication has 22 references indexed in Scilit: