Research On Cloud Computing Scheduling Strategy Based On Load Forecasting And GGA

Cloud computing is a new type of information resource service system based on the Internet.Through virtualization technology,infrastructure hardware resources are virtualized,so that massive distributed resources form a unified giant resource pool.The scheduling of user tasks is one of the core issues of cloud computing.Reasonable use of relevant resource management methods and technologies to ensure the rationality of virtual computing resource nodes,efficient allocation and use,how to achieve efficient scheduling of cloud computing resources has become an important topic of cloud computing research.In order to solve the problems of low utilization of cloud computing resources and unbalanced system load,this paper studies the cloud computing system,analyzes the characteristics of task scheduling in the cloud computing environment,and summarizes the shortcomings of scheduling strategies.By considering the possible load state of nodes in the process of task execution,a comprehensive evaluation system of virtual resource nodes in cloud computing based on load prediction is designed to generate a reasonable matrix for cloud computing task scheduling,and the task scheduling scheme based on greedy genetic algorithm is optimized to improve scheduling efficiency,shorten task completion time and optimize the overall load balancing ability ? The main innovative work of this paper is as follows:(1)This paper designs a dynamic forecasting method for the short-term load of virtual computing resource nodes in cloud computing.Aiming at the complexity,variability,time relevance and other characteristics of node load in cloud computing environment,the seq2 seq model formed on the basis of cyclic neural network(RNN)is studied.The advantages of RNN network and deep neural network in different situations are used to design and realize the load forecasting D-seq2 seq model is suitable for cloud computing resource node load discrete-time multi-step prediction.The load forecasting information of cloud computing resource node is added to the comprehensive evaluation system of cloud computing resource node,and the task scheduling matrix is formed as the basis of task scheduling.Through the experimental prediction results on the public data set Google cluster data 2011,the accuracy of the proposed d-seq2 seq model in the short-term prediction of cloud computing resource nodes is verified,which shows that the model can well adapt to the load prediction in the complex and changeable cloud computing environment.(2)A dynamic scheduling scheme of cloud computing task based on greedy genetic algorithm(GGA)is designed.On the basis of GGA,a greedy selection tree genetic algorithm(w-GGA)is proposed.The algorithm uses the method of generating greedy selection tree to calculate the probability matrix of gene selection through greedy selection tree,which is used to generate individuals in the population,so that the algorithm can obtain a better initial state A genetic template which can record the frequency of gene change and control the evolution process is used to prevent the algorithm from meaningless iteration.The d-seq2 seq model and w-GGA scheduling algorithm are written into the resource scheduling module of cloudsim cloud computing simulation platform to simulate the operation of the load forecasting model and w-GGA scheduling algorithm based on d-seq2 seq in the cloud computing environment.It is proved that the proposed scheme and algorithm can effectively improve the task completion efficiency,improve the resource utilization and load balance,and ensure the service quality of cloud applications,It has high practicability and feasibility.