Research On Cost Optimization For Request Scheduling Algorithm In Geo-distributed Datacenters

With the rapid development of virtualization technology,the construction of cloud computing platforms based on virtualization technology has become even more sophisticated.In order to complete their service deployment,more and more people in some fields are using the cloud platform.Large-scale infrastructure companies use the cloud platform to sell out different types of virtual machine rental services to earn profits.Due to the globalization of its users,the regionalization trend has become more pronounced.In recent years,infrastructure providers are more inclined to deploy their data centers in many countries and regions.To cope with the ever-increasing service requests from users all over the world,many service providers have changed the operating model of server self-sufficiency in recent years.Instead,they are deploying service on virtual machines that are leased from data centers of infrastructure providers in multiple regions.Its various types of services will not only reduce the initial server accumulation and operation and maintenance of the company,but also provide better service performance.Then,for service providers,faced with the heterogeneity of the virtual machine rental prices and user request response time in different regions,how to schedule user service requests from various regions,how to choose different virtual machine lease policies have been widely attention.In the process of saving the service provider's virtual machine rent,we must also try to ensure the user's SLA.This article uses the user request response time to measure.The balanced load rate of each data center can also make it better respond to traffic peaks and other situations,making the service more stable.The load ratio is the ratio of the processing capacity of the virtual machine used in the data center to the total processing capacity.However,the diversity of the rental prices and service capabilities of virtual machines in each data center makes the service provider's rent optimization problem more complicated.In this paper,we have established a mathematical model for mixed-integer programming that saves virtual machine rents for service providers.The goal is to minimize virtual machine rent,which takes into account virtual machine processing power and price diversity factors.In these constraints,not only the response time limit of different regions and different types of service requests is considered,but also the data center load rate balance constraint is considered.In order to solve the above problem of minimization of rents,this paper proposes a scheduling algorithm in each data center and between multiple data centers to schedule service requests and adjust the number of virtual machines to be rented,so that service providers can save virtual machine rent.It responds to unexpected situations such as traffic peaks.In this scheduling algorithm,not only include the scheduling policy for the newly arrived request,but also have the re-mapping of the service request under emergency.At the same time,this paper can collect data sets by processing the number of requests for different services in different areas of multiple time periods,and then process and analyze them,and train a predictive model to predict the number of requests for each service in the next time interval.Finally,we need to estimate the required processing capacity of the virtual machine so that we can rent the right amount of virtual machines in various regions in advance.Finally,we simulate the experimental environment,use real data sets and ARIMA prediction models to conduct experiments,use the analyzed and processed experimental data to train the model parameters,select the predicted time interval according to the prediction accuracy rate,and then collect the accumulated time periods data to predict the number of requests for the next interval.In order to verify the validity of the service request scheduling algorithm TRSA proposed in this paper,we compared the experiments with the other two algorithms and performed the experimental results on the three aspects of virtual machine total rent,virtual machine lease quantity and data center load rate balance.In the end,we can see that the TRSA scheduling algorithm proposed in this paper can ensure the quality of user service,balance the data center load ratio,and save more services,taking into account the difference between the virtual machine rental price and the processing capacity.Provider virtual machine rent.