Research And Application About Resource Scheduling In Cloud Computing System

Cloud computing is a new provision paradigm of IT resource.It is designed to provide users with shared,secure,and convenient on-demand service.Resource Scheduling is one of the fundamental technology in cloud computing.As the cloud platform scale expands,resource scheduling for cloud platform faces some challenges.Such as,how to schedule resource effectively for a task to balance the usage of each kind of resource in each physical machine in a cloud,how to schedule resource effectively under the condition that the already running tasks are not affected,how to reduce the cost of resource scheduling for the high frequency of scheduling.Focusing on resource scheduling for cloud platform,this paper puts forward a batch resource scheduling strategy based on graph theory,and applies it on the Service Platform of Enterprise-University-Research Institute Cooperation of Chongqing to achieve an efficient resource utilization and dynamic scalability.In the proposed schedule strategy,firstly,forecasts the workload of each physical machine in the cloud platform and filters out physical machines which would be with high workload in the future,then,constructs a network by using the remaining nodes and their forecasted workload information finally,batch resource scheduling is done based on this network.The main contributions of this paper are listed as follows.(1)Focusing on the challenges of resource scheduling,this paper puts forward a resource scheduling policy as follows:1)Gray dynamic forecasting model is adopted to forecast the workload,such as the utilization of CPU and memory and network bandwidth and so on,of each physical machine in the cloud platform.2)Filters out physical machines whose forecasted workload is higher than the preset threshold and constructs the remaining nodes as a network according to the forecasted workload.3)On the basic of the network constructed above,takes the problem of resource scheduling as a max-flow-min-cost problem.By solving this MFMC problem,the most suitable resource scheduling solution is found.4)Schedules tasks to suitable node according to the solution.The result of experiment shows that the scheduling strategy proposed in this paper could be able to deal with the challenges of resource scheduling in large-scale cloud platform well.(2)The Service Platform of Enterprise-University-Research Institute Cooperation of Chongqing consists of a number of subsystems.Each subsystem works independently,but there are some data sharing among them.It must be able to provide online application deployment service,allowing users to rent its hardware facilities to deploy their applications.It need to provide online website constructing service,so that users can build their own website speedy online.It must be highly scalable to allow later integration of more subsystems.According to the requirement,this paper gives the architectural and functional design of it,and realizes the design.Also,the proposed scheduling strategy is applied and achieves a good scheduling effect.