Research Of Resource Management On Cloud Data Centers

With the development of cloud computing,cloud computing in recent years has been widely used in all walks of life(such as military,education,finance,e-government,e-commerce and other industries),a large number of traditional IT systems have been moved to the cloud data centers.And with the big data,mobile Internet,Internet of Things rapid development and application,which further accelerate the basis of cloud data centers' growth on quantity and scale,it also resulted in the large-scale of modern cloud data center and its complex structure and so on.In addition,the background of the diverse needs of modern applications,resource managers in addition to care about the user reliability,performance and other quality of service,they consider energy and greenhouse reductions and other issues to reduce their own expenses,and then achieve growth of profits,which brought the cloud data centers managers a great challenge.In order to achieve the efficient,reliable and low-energy resources management,this thesis studies three aspects of cloud data centers is resource monitoring,correlation modeling and resource management.The core contributions are as follows:1)A Bionic Autonomic Nervous System(BANS)based cloud resource monitoring system is developed.Based on the design concept of BANS,a cloud resource monitoring system based on BANS was designed,which fully considered the characteristics of modern cloud data center s virtualization and large-scale.In addition,the monitoring system also has BANS autonomy,can achieve some degree of self-organization,self-diagnosis,self-healing and self-optimization,which greatly reduces the monitoring system master node load,making the monitoring system can be better suited to modern large-scale cloud data center.2)A correlation model of reliability,performance and energy consumption is proposed.Firstly,the reliability-performance and reliability-energy consumption of two correlated sub-models are established under the premise of considering the hardware random failure.Which make the performance and energy consumption of the cloud service are calculated more accurately and reasonably.T hen,through the profit model,which apply time utility function to assess the income of performance and evaluate expenses by power consumption,implementing correlation analysis of reliability,performance,and energy consumption.It establishes a model foundation for reliability,performance and energy consumption of modern cloud data center for evaluation and management.3)A self-optimization resource management framework is proposed.Considering the premise that the virtual machine(VM)is dynamic change in the demand for cloud resources,based on BANS design concept,this paper proposes a cloud resource management system based on BANS for dynamic management of cloud resources.First,the monitoring information is obtained through the monitoring system.As a basis,it will cooperate with various modules of the system to achieve a self-optimization dynamic cloud resource management framework.It relies on the virtual machine's migration to meet the dynamic requirements of the virtual machine for resources: when the physical machine overload,it can choose to move out of the virtual machine to reduce the unnecessary degradation of performance;and when the physical machine underload,it can also through the integration of VM to reduce the total number of physical machines to turn on,improve resource utilization,to achieve energy saving.4)A reliability-aware multi-data centers energy cost modeling method is proposed.With considering reliability,the system,task,scheduling,power consumption and constraint model of distributed multi-data centers are established,and the multi-data centers management problem is also formulated as total energy cost minimization with meeting the requirement of user's performance and reliability constraints.And proposed a reliability-aware algorithm which used for distributed multi-data centers energy cost optimization.Based on the reliability,the paper makes full use of the difference of electricity price in different regional data centers to the data centers' energy cost optimization.