MetaLab researcher Peter Garraghan in collaboration with Zhejiang University, China have proposed a new technique to dramatically reduce energy consumption of datacenters. A contradiction that exists within modern datacenters is that reduction of computing power can directly result in additional facility cooling power required due to hotspot formation (and vice versa). By bridging the gap in techniques and coordination between these two facets of system operation, researchers have been able to proposed a new form or scheduling algorithm which attempts to optimize Virtual Machine placement with minimal impact upon performance, leading to energy reduction by 43%.
The paper is titled “Holistic Virtual Machine Scheduling in Cloud Datacenters towards Minimizing Total Energy” and appears in IEEE Transactions on Parallel and Distributed Systems.
The authors is co-authored with Xiang Li, Xiaohong Jiang, Zhaohui Wu from Zhejiang University, and Jie Xu from the University of Leeds.