Study On Resource Scheduling Based On The Type Of Service In Cloud Computing

Cloud computing is becoming a more and more popular research field in recent years. The coming of cloud computing not only changed the technical Infrastructure of the Internet, but also have a profound impact on the entire IT industry. The data stored in the cloud data center and applications require computing power, storage and bandwidth from the data center. In a cloud computing environment, the cloud resource allocation and scheduling is one of the core technology of cloud computing platforms. Now, with the increasing demand for cloud computing, the research of resource allocation and scheduling, how to improve resource utilization of the cloud and how to reducing energy consumption has more significant significance and technical value.Based on the analysis of existing cloud computing environment and Open Stack platform, this thesis start with the perspective of the resource quantization, research the quantization method of various hardware resources in cloud environment, and finally select the benchmark test to quantified resources in cloud computing environment and build quantify resource architecture model. On the above basis, this thesis build a type of service model of virtual machine based on the needs of different virtual machines and the usage of resources, the service type of virtual machine is divided into three categories: service type of transaction processing, service type of data and IO, and service type of network traffic.When assigning a physical machine to a virtual machine, considering the different types of services of virtual machine in the distribution and migration scheduling operations, prioritizing the resource weight of the types of services in order to achieve a better quality of service and resource load level. In order to achieve the dynamic migration of virtual machines, optimize the allocation of the physical machine and migrate to another physical machine when the load changes, this thesis designs a monitoring agent mechanism to collect physical machine’s and virtual machine’s historical load data. On the basis of research on the prediction model of least square method, this thesis selected cubic polynomial model to predict the future load of physical machine, and ultimately proposed to a strategy of physical host choosing and virtual machine migration based on the physical machine’s current resources load status and future forecasting model of resource load. In the virtual machine migration strategy, according to two different types of physical machines load level, we implemented category management strategies: one migration policy is based on divergence pattern: this virtual machine migration strategy is for the purpose of reduce the load of physical machine when the load of physical machine is high; the another migration strategy is based on the aggregation mode: When the load of physical machine is low, virtual machine migration strategy reduces the energy consumption of physical machine. Finally, this thesis designed experiments in virtual machines assigned to physical host and verify the feasibility and effectiveness of this strategy when physical machine load is too high or too low.