A Performance Interference-aware Virtual Machine Placement Strategy for Supporting Soft Real-time Applications in the Cloud

It is a standard practice for cloud service providers (CSPs) to overbook physical system resources to maximize the resource utilization and make their business model more profitable. Resource overbooking can, however, lead to performance interference between the virtual machines (VMs) hosted on the physical resources thereby causing performance unpredictability for applications hosted in the VMs. As applications with soft real-time requirements get increasingly hosted in cloud environments, performance jitter caused due to resource overbooking will be unacceptable to these applications. Addressing these conflicting requirements needs a careful design of the placement strategies for hosting soft real-time applications such that the performance interference effects are minimized while continuing to allow resource overbooking. These placement decisions cannot be made offline because run-time workload changes and satisfying the priorities of collocated VMs may require VM migrations, which requires an online solution. This paper presents a machine learningbased, online placement solution to this problem that learns from a publicly available trace of a large data center owned by Google. Our approach first classifies the VMs based on their historic mean CPU and memory usage, and performance features. Subsequently, it learns the best patterns of collocating the classified VMs by employing machine learning techniques. These extracted patterns are those providing the lowest performance interference level on the specified host machines making them amendable to hosting soft real-time applications while still allowing resource overbooking. Keywords-virtual machine placement, cloud computing, deployment algorithm, performance interference and QoS.