Energy-saving analysis of Cloud workload based on K-means clustering

With the development of cloud infrastructure services, IaaS(Infrastructure as a Service) study on energy-saving technology has been attracted more and more attention. IaaS platform providers can provide high performance service for the users. Meanwhile, how to save the energy cost of the cloud platform must be considered without violating the Service Level Agreement(SLA). The overload and underload are two running statuses of physical machine(PM), the former will cause the possibility of SLA violation, while the latter will cause the low utilization rate of PM´s resources, causing additional energy consumption. This paper proposes a model of workload characteristic based on K-means clustering analysis, using Google workload trace data set, which is the basis of virtual machine(VM) migrating when PM has been underloading or overloading. The establishment of workload characteristic model can present the demand of system resources in real time so that VM scheduling strategies carry out efficiently.