Evolutionary multiobjective optimization for green clouds

  • Authors:

  • Dung H. Phan;Junichi Suzuki;Raymond Carroll;Sasitharan Balasubramaniam;William Donnelly;Dmitri Botvich

  • Affiliations:

  • University of Massachusetts Boston, Boston, MA, USA;University of Massachusetts Boston, Boston, MA, USA;TSSG, Waterford Institute of Technology, Waterford, Ireland;TSSG, Waterford Institute of Technology, Waterford, Ireland;TSSG, Waterford Institute of Technology, Waterford, Ireland;TSSG, Waterford Institute of Technology, Waterford, Ireland

  • Venue:
  • Proceedings of the 14th annual conference companion on Genetic and evolutionary computation
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

As Internet data centers (IDCs) have been increasing in scale and complexity, they are currently a significant source of energy consumption and CO2 emission. This paper proposes and evaluates a new framework to operate a federation of IDCs in a "green" way. The proposed framework, called Green Monster, dynamically moves services (i.e., workload) across IDCs for increasing renewable energy consumption while maintaining their performance. It makes decisions of service migration and placement with an evolutionary multiobjective optimization algorithm (EMOA) that evolves a set of solution candidates through global and local search processes. The proposed EMOA seeks the Pareto-optimal solutions by balancing the trade-offs among conflicting optimization objectives such as renewable energy consumption, cooling energy consumption and response time performance.