An Online Virtual Machine Placement Algorithm in an Over-Committed Cloud

Cloud computing provides users and enterprises shared pools of resources to store and process their data. Virtualization is one of the technologies used in cloud computing to separate the underlying hardware resources to create virtual machines (VMs). VM placement is the process of choosing the best PM for newly created VMs. Since different users have a variety of resource requirements, how to deploy VMs becomes a challenging problem in cloud management. Most of the existing works only consider maximizing the resource utilization of PMs without taking the overcommitment issue into account, which can cause PM overloading and degrade VM performance. In this paper, we propose an algorithm, called Min-DIFF, that can balance the usage of resources along multiple dimensions and reduce the risk of PM overloading in a threshold-based placement strategy. To better evaluate the performance of Min-DIFF, our simulations are driven by both real-world workload traces and datasets we generated. Extensive simulation results show that Min-DIFF achieves a significant improvement over the existing approaches in balancing the utilization of multi-dimensional resources and reducing the risk of PM overloading.