Research On Scheduling Strategy Of Cloud Computing Based On Application Type

To increase the utilization efficiency of cloud resources,study of scheduling and allocating cloud resources has attracted a lot of attention.However,the traditional resource scheduling is to try to optimize and improve load balancing algorithms by forecasting the demand for resources based on historical data.This method needs a lot of historical data which costs a long time to collect.To improve the scheduling and execution efficiency of applications,this paper studies the scheduling strategy based on application types.Through the analysis of the application of cloud computing,an application recognition model is proposed to classify the applications.And the corresponding scheduling strategies are put forward based on the types of applications to realize the scheduling for the specific applications.And on this basis,the scheduling model for hybrid applications is established to implement the scheduling for hybrid application.Finally,the dynamic adjustment strategy of virtual machine based on application load is proposed.The main contributions of this paper are as follows:Firstly,we have studied the classification problem in the cloud computing environment.Analyze the characteristics of CPU,IO and network intensive applications according to the resource consumption of different types of applications.According to the characteristics of each type of applications,we draw out the main parameters,which can be used as the features to classify different type of applications.Using these features an application recognition model is proposed.Secondly,we study the scheduling strategy of applications.For the typical applications such as CPU,IO and network intensive applications,we put forward the corresponding scheduling strategies based on the types of applications to improve the execution efficiency of applications.For the hybrid applications,according to the characteristics of the application and the scheduling model of typical applications,a scheduling framework for hybrid applications is proposed.And it improves the execution efficiency of the hybrid applications effectively.Thirdly,to effectively utilize the resources of virtual machine(VM),we propose a dynamic adjustment strategy for VM based on load forecasting.The strategy predicts the system resources usage and application service response time,and finds the key resource which affects the system state by comparing the size of the key resource factors.After gaining the key resource,scale the key resource appropriately.If the system is in low load state,it is appropriate to reclaim the key resource and reduce the waste of resources.If the system is overloaded,increase the key resource dynamically which is the system bottleneck to achieve the goal of improving the system performance.Fourthly,in order to verify the correctness of the application recognition model,the scheduling strategy of typical and hybrid applications and the virtual machine dynamic adjustment strategy,extensive experiments have been done.We make analysis of the experimental results of each model or strategy,and the experimental data are summarized.Experimental results show that the model and scheduling strategies are correct,and they can be used efficiently to classify different type of applications which can serve as the basis of effective use of cloud resources and decrease the energy consumption.