Multiple Resource Allocation Method Based On Dominant Resource Fairness In Cloud Computing

With the development of Cloud Computing and Big Data processing,domestic and foreign enterprises and service providers urgently need to deploy multiple computing frameworks to share resources and data within a cluster in one common cloud environment.The problem of resource allocation in multi-resource environments has been extensively studied.Aiming at the load balancing problem of the Dominant Resource Fairness algorithm in multi-resource fairness distribution environment,this thesis proposes an algorithm to improve cluster load balance of resource allocation.This method has the following innovations:1.Using multi-objective optimization improves K-means clustering method.Using multi-objective optimization method NSGA-II optimize the initial clustering center in K-means clustering,to avoid fall into local optimum in the process of K-means clustering.It makes the classification result more accurate.2.MOCDRF algorithm for improving cluster load balancing is proposed.Using multi-objective optimization of K-means clustering method analyzes the resource utilization of nodes in cluster.According to given the number of centers in clustering,nodes are clustered by various resource utilizations on each node in the cluster.The node resources with small resource utilization are preferentially allocated,which makes the overall load of cluster more balanced.3.Improved the fairness of the Dominant Resource Fairness algorithm.In the process of resource allocation,the user priority value is calculated by the time waiting factor and the dominant share to improve the fairness of resource allocation.It makes the allocation scheme more reasonable,and improves user satisfaction.Based on the above improvements,this thesis implements a multi-resource allocation tool to improve load balancing.By analyzing the experimental results,Dominant Resource Fairness algorithm based on multi-objective clustering can effectively improve the rationality of resource allocation and make the cluster load more balanced.