Output details
11 - Computer Science and Informatics
University of Leeds
Improved energy-efficiency in cloud datacenters with interference-aware virtual machine placement
<02>Virtualization in Clouds can result in contention for the underlying physical resources, adversely affecting performance and energy-efficiency. Previously, the dynamic relationship between workload co-allocation, performance, and energy-consumption was unknown; this paper is the first worldwide to identify and verify this relationship, using real data from Google. Using our findings, we develop a novel method for controlling virtualization contention and significantly increasing energy-efficiency. Voted best paper by 200+ participants (IEEE-ISADS'2013). International impact includes keynote address (IEEE/CIE-CSC'2013), follow-on publication (IEEE-CloudCom'2013), EU efficient-Clouds project, and new collaborations (Prof Jose Fortes, Florida; Michael Walker, AIMES UK). Commercialisation (patent application & spinout) is on-going.