Research Of Resource Scheduling In Cloud Environment

In recent years,cloud platforms have attracted a lot of attention due to their unique resource virtualization,convenient access,transparent billing and low prices.At the same time,the surge in user usage has been accompanied by the increase in the actual operating pressure of cloud platforms In a cloud environment,the main factors affecting the performance of the cloud platform in resource scheduling and allocation,therefore,it is of great significance to study and design efficient and reasonable resource scheduling schemes.This paper analyzes the process of the two parties involved in the cloud platform,from task submission to resource allocation,and gives the actual optimization plan.The main tasks are:(1)Establish a two-way auction model.According to the expectations of user needs and the expectations of cloud service providers,build a new two-way auction model.Due to the diversity of cloud users' target tasks,different cloud users have different configuration choices for cloud servers.Even for cloud servers with the same configuration,the expected expenses of different cloud users will be different.When determining the auction result,the model not only takes the bid expectations of both parties as the bidding standard,but also considers the waiting time and priority of participating in the bidding task,and proposes the concepts of queuing influence factor and penalty system.The queuing influence factor adjusts the auction priority according to the waiting time of both agents when participating in the auction and the number of waiting rounds for participating in the auction to ensure the timeliness of completing the task.The penalty system can prevent malicious auction bids and prevent tasks from waiting to be stuck,which can not only improve the user experience,improve the efficiency of completing tasks,but also ease the load on the platform.The simulation experiment is carried out by constructing relevant constraint functions to verify the reasonable feasibility of the model.(2)Design cloud virtual machine dynamic migration scheduling strategy.After completing the resource allocation process according to the two-way bidding model,when the cloud service provider receives the allocated corresponding order,the cloud virtual machine migration strategy is optimized at the resource scheduling level.Use dynamic criteria to determine whether the current state of the cloud virtual machine is underloaded or overloaded,and then make a matching dynamic migration strategy based on the current actual state.This strategy solves the problem of cloud service providers' own cost and energy waste,and reduces the cost of cloud service providers from a cost perspective.(3)Early warning of cloud server operation failures through data mining.The cloud server failure warning is mainly aimed at improving the user's experience of using the cloud platform,and at the same time ensuring the smooth operation of the cloud user during the use of the cloud server.In the process of users using cloud servers,a cloud server failure data analysis scheme based on improved FP-Growth algorithm is proposed,combined with the real recorded data values generated when users use cloud servers,to monitor the current running state of cloud servers in real time and predict possible failures Situation,early warning of failures to users to improve user experience.