Genetic Algorithm Based Virtual Machine Consolidation Strategy

With the rapid development of cloud computing,the amount of modern cloud data centers is expanding,and the problem of high energy consumption needs to be solved urgently.The efficiency of the virtual machine consolidation(VMC)strategy is the core of the data center energy consumption,so studying the VMC is an effective way to solve this problem.At present,the core of VMC is to transfer the virtual machines(VM)on the low-load physical hosts to achieve zero load and thus make them sleep,so as to reduce energy consumption.However,excessively efficient VMC can easily cause the running physical host to fail to cope with the load spike.When the host load suddenly increases,the dormant physical host needs to be turned on,and the resulting time loss will seriously affect the quality of service.Therefore,it is of great practical significance to find an efficient VMC strategy,which requires minimizing energy consumption while ensuring service quality.Through in-depth analysis of existing VMC strategies,this project studies and proposes two efficient VMC strategies.The research content is as follows:Through the in-depth study of the existing overload detection method of physical host and how to ensure the long-term stability of the physical host load after VM migration,a resource matching-based VMC strategy is proposed.This strategy detects whether the physical host is overloaded.First,use the Markov chain to predict the future load value of each physical host in the data center,and then compare the current load value and predicted value with the preset threshold of the physical host to determine whether the physical host is overloaded.To ensure the long-term stability of the physical host load after the VM is migrated,firstly,the workload of the physical host is quantified according to the historical data of the VM resource request and the history of the available resources of the physical host.The core of this method is to quantify the resource matching degree between the VM and the physical host.This resource matching degree represents the remaining available resources of the physical host requesting the VM.The degree of resource satisfaction is solved by gradient descent method,and then the reallocation between virtual machine and physical host is completed according to resource matching degree.Experiments show that this strategy can ensure the long-term stability and quality of service of the physical host load,while reducing energy consumption and the number of VM migration.Based on the above research results,in order to further optimize the data center energy consumption,a genetic algorithm based VMC strategy is proposed.This strategy considers the virtual machine placement problem as a multi-objective optimization problem with energy consumption and multi-resource constraints,and then solves it with an improved genetic algorithm.For the genetic algorithm,optimization is carried out in the aspects of selection,crossover,mutation and fitness function design.The genetic evaluation parameters are defined in the crossover and mutation stages,and the first-time adaptive algorithm is combined with the descending order.Finally,the optimal search solution is improved.Efficiency and maintain population diversity.Experiments show that this strategy has a significant improvement in reducing energy consumption and improving service quality.