Low SLA violation and Low Energy consumption using VM Consolidation in Green Cloud Data Centers

Virtual Machines (VM) consolidation is an efficient way towards energy conservation in cloud data centers. The VM consolidation technique is applied to migrate VMs into lesser number of active Physical Machines (PMs), so that the PMs which have no VMs can be turned into sleep state. VM consolidation technique can reduce energy consumption of cloud data centers because of the energy consumption by the PM which is in sleep state. Because of VMs sharing the underlying physical resources, aggressive consolidation of VMs can lead to performance degradation. Furthermore, an application may encounter an unexpected resources requirement which may lead to increased response times or even failures. Before providing cloud services, cloud providers should sign Service Level Agreements (SLA) with customers. To provide reliable Quality of Service (QoS) for cloud providers is quite important of considering this research topic. To strike a tradeoff between energy and performance, minimizing energy consumption on the premise of meeting SLA is considered. One of the optimization challenges is to decide which VMs to migrate, when to migrate, where to migrate, and when and which servers to turn on/off. To achieve this goal optimally, it is important to predict the future host state accurately and make plan for migration of VMs based on the prediction. For example, if a host will be overloaded at next time unit, some VMs should be migrated from the host to keep the host from overloading, and if a host will be underloaded at next time unit, all VMs should be migrated from the host, so that the host can be turned off to save power. The design goal of the controller is to achieve the balance between server energy consumption and application performance. Because of the heterogeneity of cloud resources and various applications in the cloud environment, the workload on hosts is dynamically changing over time. It is essential to develop accurate workload prediction models for effective resource management and allocation. The disadvantage of VM consolidation process in cloud data centers is that they only concentrate on primitive system characteristics such as CPU utilization, memory and the number of active hosts. When originating their models and approaches as the decisive factors, these characteristics ignore the discrepancy in performance-to-power efficiency between heterogeneous infrastructures. Therefore, this is the reason that leads to unreasonable consolidation which may cause redundant number of VM migrations and energy waste. Advance artificial intelligence such as reinforcement learning can learn a management strategy without prior knowledge, which enables us to design a model-free resource allocation control system. For example, VM consolidation could be predicted by using artificial intelligence rather than based on the current resources utilization usage