Research On Cloud Computing Scheduling Algorithm Based On QoS Constraints

Recently, in industry and academia vigorously promoting, cloud computingtechnology obtains the rapid development, and a large number of cloud computingsystems have been applied. With the expansion of the scope of cloud computingapplication, the number and types of users is increasing, and the special needs ofcloud computing resources is also gradually increasing. Cloud computing as abusiness model, its commercial character of the decision more cloud computingsystems need to focus on the user demand for services, to provide users with highquality service. The scheduling technique as one of cloud computing key technology,it is mainly responsible for the user’s job assigned to the appropriate resources. Butthe existing scheduling algorithms have not fully consider the different needs of theresources, so they can not be good for different users to provide to meet therequirements of the resources.For the QoS needs of cloud computing and the shortcomings of the originalcapacity scheduling algorithm of cloud computing without considering user diversity,this paper presents the capacity scheduling algorithm based on QoS constraints.Based on the guarantee to provide users with simulation independent computingcapability, the algorithm can match resources and tasks by QoS parameter generatingvector, and distinguish between different user service quality requirement, provideusers with its demand for resources. It also changes the original two class schedulingto three class scheduling, in order to reduce the scheduling overhead of theintroduction of QoS matching.In addition, this paper research on scheduling mode of Hadoop system of cloudcomputing in the main stream. Based on the problem of a single master schedulernode Master overloaded, this paper introduces AssistMaster node. This node isresponsible for scheduling in the matching calculation, with the Master node tocomplete the job scheduling to reduce the computational load of the Master node. Itchanges the original scheduling mode to User-Master/AssistMaster-Worker mode,and enhances the overall operating efficiency of the cloud computing system. Finally, in CloudSim cloud computing simulation platform, the proposedscheduling algorithm and scheduling model experiments verified the feasibility of thescheduling algorithm as well as service quality satisfaction guarantee. Schedulingload is balanced, and overall operating efficiency of cloud computing systems isimproved.