It is a multiobjective optimization algorithm which is based on the improved GA that utilizes Pareto theory to achieve the selection, crossover, and mutation operators of GA towards multiple objectives and thus to find out the set of Pareto optimal solutions of the proposed multiobjective problem by the evolution of the population. After obtaining the set of Pareto optimal solutions, in order selleck Gemcitabine to get the final migration policy, we do not randomly select a solution but utilize the mathematical statistics theory to obtain the final solution of location selection of live VM migration by using a probability wheel for each VM. Specifically, we have employed a novel policy which is used to generate the initial population and defined a Pareto constraint dominance relation towards the proposed constrained problem to compare two solutions in the improved GA-based approach.
The nondominated sorting policy and density estimation policy used for the population has been presented to make each individual of the population obtain fitness values in each generation. We have employed the tournament selection operator for the selection operator of GA. For the crossover and mutation operators, we have designed the arithmetic crossover operator and the dynamic nonuniform mutation operator. Besides, the (�� + ��) selection policy, which makes elitism that helps in achieving better convergence be introduced into the proposed MOGA-LS approach, is employed to generate the next population.3.2.
Design of the MOGA-LS ApproachThe MOGA-LS approach is an algorithm based on multiobjective GA achieving a live VM migration policy for minimizing the incremental power consumption caused by migrating these migrant VMs onto their target hosts and making the load of cloud data center balanced after migrating under the constraint of maximizing the performance that the number of success of live VM migration events is maximized. To achieve a multiobjective GA, we have introduced the concept of Pareto optimal solutions into GA and designed a constrained Pareto dominance method to evaluate the individuals of population and assign fitness values to them as well as designing GA’s genetic operators including selection, crossover, and mutation operators as well as the policy of generating the new population.3.2.1. Problem Formulation We now formulate the proposed problem of migrating n VMs onto m hosts.
Its solution can be represented by an n dimension of vector, each element of which denotes the target host of the migrant Carfilzomib VM which its location represents. We assume that there are m available hosts in the resource pool of a cloud data center and the hosts are heterogeneous and dynamic while using space shared allocation policy. The hosts change their state dynamically according to the load. Similarly, each VM is associated with the required computing resource.