GreenPOD: Leveraging queuing networks for reducing energy consumption in data centers

Abstract

Reducing the energy consumed in Clouds is becoming an increasingly challenging research direction due to the exponential growth of services hosted there and delivered. The big concern of data center providers is to find a trade-off between the offered resources and the current traffic load. GreenPOD aims to reduce the consumed energy as well as to satisfy pre-defined response time for fat-tree data center networks by using queuing theory based on mathematical model. GreenPOD considers different activation thresholds based on request queuing to switch ON/OFF different POD which is formed by a certain group of servers when needed. Our results show that GreenPOD outperforms previous energy-aware queuing theoretical model. In respect of a moderately demanded system (resp. highly demanded system), GreenPOD saves up to 61.72% (resp. 47.75%).