Research On Resource Scheduling Algorithms Based On The Multiobjective Optimization In Cloud Computing

Cloud computing is considered as a service model to provide the services for the users' demands.One of its key techniques is the virtualization,in which the physical computing resources can be effectively allocated by the resource scheduling to provide the required services.However,due to the increasing demands for the computing resources,the scale of the datacenters becomes larger and it is more complex and difficult to manage and allocate the resources.It becomes an issue to improve the resource utilization and to reduce the energy.Hence,it is beneficial to propose an adaptive resource scheduling strategy to improve the resource utilization and reduce the energy consumptions simultaneously.In this paper,to improve the utilization and reduce the energy,a further research is made on the the multiobjective optimization and adaptive resource scheduling strategy.The main contents are listed as below.First,an adaptive resource scheduling strategy on the performance analysis is proposed.By analyzing multiple key indicators,the performance threshold is determined by using the entropy method based on the survivability.Then,to meet the users' demand,a rapid adaptive resource scheduling strategy is proposed.When it is in underprovisioning,the servers would be scaled up by the PM level or VM level.When it is in overprovisioning,the spare servers would be shut down to save the energy consumption based on the WMA method.The experiment has shown that the proposed adaptive scheduling algorithm would rapidly allocate the resources to meet the users' demand.Secondly,a three-phase resource optimization approach is proposed.Since the demands are fexible and the burst loads happen,a single researvation plan can't meet the dynamic demands.A three-phase resource optimization strategy is presented,including the reserved plan,predictive technique and resource consolidation strategy.In the reserved phase,a reserved plan is presented by using the stochastic integer linear programming modeling.Based on the reserved plan,a hybrid predictive technique is proposed to determine the next demand in the future,which includes a double exponential smoothing and WMA method.In the resource consolidating phase,a simple resource consolidating strategy is presented,which would reduce the energy by shutting down the spare machines.The experiment has shown that the proposed three-phase algorithm would rapidly allocate the managed resources by minimizing the overheads.Thirdly,an improved load balancing strategy on the PSO algorithm is proposed.In the resource scheduling,not only the energy consumption should be focused,but also other objectives should be considered.Additionally,some main problems exist,such as how to determine the hotspots and solve the VM placement problem.Hence,a double objective load balancing strategy on the PSO algorithm is proposed to improve the utilization and reduce the energy consumption.In the load balancing scheduling,based on the entropy method,an approach to determine the hotspots is proposed by analyzing the key indicators,such as the CPU and memory.Then,by using the TOPSIS method,a double objective PSO algorithm is proposed.It can efficiently solve the VM placement problem by minimizing the energy and maximizing the utilization.The experiment has shown that the proposed load balancing strategy on the improved PSO algorithm would rapidly locate the position and make the load balancing effectively.Fourth,a multiobjective consolidating strategy on the ACO algorithm is proposed.In the resource scheduling,some main problem existed,such as when to migrate,how to migrate and where to migrate.Hence,a multiobjective consolidating strategy on the ACO algorithm is proposed,which would effectively reduce the energy consumption.In the resource scheduling,by considering multiple performance indicators,a score approach on the GRA method is proposed to determine the hotspots or underprovisioning hosts.In addition,by considering the CPU utilization and memory utilization,a selected algorithm is presented to determine the migrated machines based on the euclidean distance method.Then,to solve the VM placement problem,a multiobjective ACO algorithm is presented by using the Pareto theorem.The experiment has shown that the proposed hybrid algorithm would effectively consolidate the resources.By comparing with other algorithms,it can efficiently avoid the SLA violation and reduce the energy consumption.