Research On Cloud Resource Scheduling Strategy Based On M/M/n Queuing Model

With the rapid development of cloud computing, users are demanding more. The cloud computing center is becoming more and more scalable, which is followed by high energy consumptions and severe energy problems. Therefore, it’s urgent to carry out energy-saving technology researching under cloud computing environment. The purpose of cloud computing is to share resources in practice, and resource scheduling is one of its core problems. As a resource scheduling system requirement, energy consumption index is influenced not only by server performance, but also by application performance. There is not an optimal solution for resource scheduling problem yet. On the one hand, it’s NP-hard. On the other hand, various conditions, application requirements and scheduling objective also restrict the solving process. Not mention the energetic and environmental problems brought up by IT cost, so how to weigh the relationship between energy saving and application performance in the process of resource scheduling is a problem worthy of further study.This thesis analyzes the process of resource scheduling in cloud computing environment in the aspects of scheduling model, scheduling basis, scheduling objective and scheduling strategy. First of all, this thesis models on the cloud computing system which is based on M/M/n queuing model, and makes mathematical analysis on the performance of the model which is based on the regulation of task reaching and service under the cloud computing environment. Thus the resource scheduling model of cloud computing system is determined. Then, according to the ideas of the bionics, this thesis proposes and designs a biomimetic autonomous monitoring system. It provides scheduling basis in subsequent resource scheduling strategy research which is based on the resource monitoring data obtained by this system. And this thesis sets up the energy consumption model of cloud computing system, measures the energy consumption parameters of the specific server, and then determines the scheduling objective of lower energy consumption and less service time.Besides, cluster load classification scheduling policy is proposed according to the resource monitoring data obtained by biomimetic autonomous monitoring system combined with the energy value of the server under different load calculated by the energy consumption model. Cluster load classification scheduling policy divides the cloud computing system into these types: empty load cluster, low load cluster, normal load cluster and high load cluster according to the load on the server being monitored. In order to make the system in normal load condition in most of the time, the system takes different scheduling policies for different categories of cluster. Eventually the requirements of load balancing are met, while reducing energy consumption and service time.Finally, this thesis compares among the average response time, load on the server, system average power, system average energy consumption by using the widely used Min- Min algorithm and cluster load classification scheduling algorithm, which indicates that the cluster workload classification algorithm is practical and effective in reducing energy consumption and task service time.