Cloud Data Center Resource Dynamic Adjustment Method And System

The cloud data center provides basic platform on support of big data,Internet of Things,artificial intelligence and so on.The cloud data center shows characteristics and needs on diversity of user requirements,parallelization of task execution,and dynamic resource allocation.Dynamic resource allocation and adjustment is an effective way to utilize cloud data center resources and ensure quality of service.This thesis focuses on the dynamic adjustment of cloud data center resources,realizing user demand and real time optimized utilization of environment self-adaptation resources.Following are detailed content of research.The dynamic resource adjustment method based on Simplex and Smart Hill-climbing algorithm has problems on iterative frequency,slow convergence rate and low efficiency,and it is difficult to respond quickly to changes in application requirements.Associate the prior knowledge of application execution,build resource optimization allocation template based on reinforcement learning(RL)and adaptive resource dynamic adjustment method,realize rapid mapping of cloud data center resources and application requirements,and improve quality of service of cloud data center services on dynamic resource demand.Experiments show that compared with the Smart Hill-climbing algorithm,the proposed method resource dynamic allocation delay is reduced by 22%.Aiming at the low efficiency of the learning process based on RL-based resource dynamic adjustment method,the action selection method based on prior knowledge is proposed to accelerate the learning process and improve the learning process efficiency.Aiming at the problem that the RL-based resource dynamic adjustment method has large search space,the Simplex algorithm is used to obtain the optimal configuration set.The Simplex-based state space reduction algorithm and the Simplex-RL hybrid resource dynamic adjustment method are proposed to reduce the search space and improve the efficiency of the algorithm.Experiments show that compared with the Smart Hill-climbing algorithm,the proposed method resource dynamic allocation delay is reduced by 30%.Compared with the RL-based resource dynamic adjustment algorithm,the proposed method resource dynamic allocation delay is reduced by 6%.Based on the research method,the cloud data center resource dynamic adjustment system is designed and realized.The system has the functions of monitoring application load,associating the prior knowledge of application execution,learning resource optimization allocation template,resource dynamic adjustment and other functions.These functions meet the demand of real-time and quality of service for resource adjustment.The system is applied to the cloud computing platform of the cloud data center and verifies the effectiveness of the research method.