Power-aware scheduling on clouds using active cooling control and DVFS

Author(s)

Daniel Guimarães do Lago

• Document type: Master's Dissertation
• Press: Campinas, SP.
• Institution: Universidade Estadual de Campinas (UNICAMP). Instituto de Computação
• Defense date: 2012-11-29

Examining board members

Edmundo Roberto Mauro Madeira; Fabio Luciano Verdi; Nelson Luis Saldanha da Fonseca

• Advisor: Edmundo Roberto Mauro Madeira

Abstract

This work presents algorithms that operate in the scheduling process of virtual machines in cloud computing, based on concepts of Green Computing. These algorithms focus on energy consumption minimization, and also provide quality of service based on virtual load execution priority. To achieve energy consumption minimization, some features are used, such as underutilized server shutdown, migration of server loads that operate below a certain threshold, DVFS, and Fan Control. The choice of which server will receive the virtual loads is obtained through energy efficiency maximization, given by the ratio MIPS to power consumption per server. Furthermore, the algorithms of this work also act on the submission of loads with priorities, allowing high priority loads to run in lower times. Results from simulations showed that the developed algorithms, when confronted against other studied algorithms, can improve energy consumption in clouds composed of heterogeneous data centers, tying with the best algorithms in the case of homogeneous data centers, also providing quality service (AU)

• FAPESP's process: 10/14592-9 - A scheduling algorithm for cloud computing using Green Computing concepts
• Grantee: Daniel Guimarães do Lago
• Support Opportunities: Scholarships in Brazil - Master