Resource Scheduling Methods And Key Technologies Supporting Green Cloud Computing

In recent years, with the rapid development of cloud computing, the hardware scale continues expanding in cloud data centers, which leads to huge amount of energy consumption. The energy dissipation problem has received extensive attention from both industry and academia, and how to realize green cloud computing is emerging as a critical issue to promote the healthy development of cloud computing industries. Virtualized resource scheduling is an important technical means to promote resource allocation and management for cloud data centers, and it affords powerful technical support to realize green cloud computing. On one hand, virtualized resource scheduling can ensure the technical features of cloud computing, including on-demand access and pay-as-you-go pricing, and it also can meet the elastic needs of users for QoS (Quality of Service). On the other hand, virtualized resource scheduling can support the implementation of green cloud computing effectively, and achieve energy savings in the cloud data centers, from the perspective of energy optimization.Technically, virtualized resource scheduling provides feasible ways for the implementation of green cloud computing. However current researches about green cloud computing mainly focus on energy saving, but they often neglect the relationship between QoS and energy consumption during resource scheduling, including the effect between QoS enhancement and energy consumption, as well as the control relationship between large-scale live migrations and personalized resource requests. Thus resource scheduling supporting green cloud computing still suffers from the following challenges.1) Energy has not been taken into consideration in traditional QoS evaluation system; 2) in order to achieve energy savings, the cloud data centers suffer from the complicated and dynamic migration requests from virtual machines (VMs), and it is necessary to balance the energy consumption and the performance degradation due to VM migrations; 3) as there are large amounts of applications with various resource requirements, it is essential to design energy-efficient resource scheduling strategies for the applications with personalized features.In view of these challenges, we propose our solutions for resource scheduling supporting green cloud computing. More specifically, our contributions are summarized as follows.1) In order to achieve the goal of energy saving and emission reduction in the cloud data centers, a resource scheduling framework supporting green cloud computing is proposed. This framework consists of five layers, i.e., hardware layer, virtualization layer, scheduling method layer, scheduling technology layer and application implementation layer. Concretely, the hardware layer refers to the hardware infrastructure, refrigeration and air-conditioning equipment, as well as ventilation installations. The virtualization layer provides several types of VM instances for different kinds of applications and supporting techniques for data center management. According to the VM instances and the supporting techniques provisioned by the virtualization layer, the method layer provides energy-efficient resource scheduling for QoS enhancement. Then the technology layer is employed to optimize the scheduling processes in the method layer, and a corresponding resource scheduling strategy for trade-offs between energy and performance degradation is identified in this layer. Finally, in the application implementation layer, resource allocation and scheduling are conducted for the applications with personalized resource requirements and scheduling features.2) In order to satisfy the requirements for QoS enhancement of cloud consumers, a QoS-enhanced resource scheduling method supporting green cloud computing is proposed. Concretely, an energy consumption model in cloud environment is presented to analyze all aspects of the energy consumption during task executions. Furthermore, an energy-aware VM scheduling policy is proposed to make the most use of PMs with lower energy consumption rate. Some VMs on the PMs with high energy consumption rate are moved to the PMs with lower energy consumption rate. After this migration operation, the empty PMs are shut down or set to sleeping mode to achieve energy savings. Energy savings lead to cost savings for cloud data centers, which can benefit users with discount prices. Then, in order to reduce the execution time of some tasks, an execution time aware VM scheduling policy is designed to make best use of the PMs with higher performance, and some VMs on the PMs with lower performance are moved to the PMs with higher performance. After this migration operation, the execution time of certain tasks can be reduced to a great extent, and users can experience the dynamic QoS optimization directly. Finally, migration pruning based VM scheduling is conducted to optimize the VM migrations and realize the energy-efficient VM scheduling for QoS enhancement.3) To achieve energy optimization and mitigate VM performance degradation due to VM migrations, a resource scheduling technology for balancing energy and performance is proposed. Concretely, to quantify the effects of energy consumption and VM performance degradation by VM migrations, an energy consumption model and a performance model of VM migrations are constructed respectively. Furthermore, normalization processing of energy consumption and performance degradation is carried out, based on SAW (Simple Additive Weighting) and MCDM (Multiple Criteria Decision Making) techniques. Then through the analysis of VM distributions, a heuristic search for VM migration strategies is designed. Finally, according to the utility functions by the normalization processing, the optimized VM migration strategy is identified to realize the VM scheduling for trade-offs between energy consumption and performance degradation.4) To verify the effectiveness and the extensibility of our proposal (i.e., the resource scheduling method and the resource scheduling technology supporting green cloud computing), the analysis and research of scientific workflow applications supporting green cloud computing are conducted. Concretely, the resource requirements are analyzed according to the personalized features of scientific workflows, and all aspects of energy consumption generated during scientific workflow executions are also analyzed in detail. Furthermore, a resource utilization table is defined to trace the resource usage for the subtasks in the scientific workflows, and monitor the resource utilization of PMs in the cloud data centers. Then static resource allocation is conducted according to the personalized resource requirements to maximize the resource utilization. Finally, to achieve high energy efficiency, the live migration technique is leveraged to dynamically schedule VMs during scientific workflow executions.