Research On Performance Optimization Of Computing Model In The Cloud Computing Environment

Cloud computing has brought a new revolution into the IT industry. The keytechnologies of cloud computing include virtualization technology, distributed filesystems, parallel computing technologies, NoSQL database technology, and so on.Hadoop is an open source cloud computing platform synthesizing a variety of cloudcomputing technology, mainly consists of HDFS file system and MapReducecomputation model.Hadoop can storage and process massive data by clusters composing of multiplecomputers, and has the characteristics of scalability, fault tolerance and efficiency.The ideas of HDFS file system and MapReduce programming model come fromGoogle. Hadoop is in a constantly mature stage. To Hadoop system framework, manyresearchers have analyzed and proposed a variety of improvement measures. Inaddition to architecture, Hadoop configuration has an import effect on the systemperformance. This paper mainly studies the method of how to configure Hadoop inorder to improve the performance of the Hadoop system.Hadoop manages the whole system through the configuration. Hadoopconfiguration optimization problem is a complex combinatorial optimization problem.There are dependencies and constraints among the various attributes. Unreasonableconfiguration will lead to competition for resources, and reduce the overall systemperformance. Intelligent algorithm can solve combinatorial optimization problems.Based on Intelligent algorithm, configuration optimization strategy specific to certainapplication first be proposed. Considering the actual usage of cloud computingplatforms, a system configuration optimization model specific to multi-applicationenvironment is proposed.For specific applications, through genetic simulated annealing algorithm, takeeach configuration scheme as a chromosome, then constantly select, cross, mutate.And combine simulated annealing principle to control the survival of the newchromosome and the number of iterations of the algorithm, to find out the optimal system configuration scheme. Experimental results show that this configurationmethod can effectively improve the operating efficiency of the job, and the wholegroup performs perfect at the end of the iteration.For specific applications, through the chaotic particle swarm algorithm, take eachconfiguration scheme as particle’s location, by the constant motion of the particles tofind the optimal configuration. To make algorithm can search both in global and inlocal, and also can avoid falling into local optimum prematurely, a new processmethod is proposed by combing the particle swarm algorithm, the improvement ofweighting coefficient and maximum speed of processing methods and chaosalgorithm. Experimental results show that this method also can significantly improvethe efficiency of the system. The optimal configuration derived from experiment hashigh stability. The method is simple and the number of iterations is very small.For practical operation scene, the system needs to handle all types of jobs, soconsidering a variety of applications, an adaptive system configuration optimizationmethod is proposed. According to jobs’ characteristics of the cloud computingplatform, adaptive algorithm divides the jobs to two cases in the process of systemconfiguration optimization. When the jobs’ types of operating system are similar,according to characteristics of the whole group of the genetic simulated annealingalgorithm all move towards the optimal solution, based on genetic simulatedannealing algorithm, a real-time optimization configuration model is proposed. Whenthere are remarkable differences among the jobs of the system, anexperience-reference model is proposed, configuring the system according todatabases queries. If the job type does not exist in the database, considering thesimplicity and rapidity of the chaotic particle swarm algorithm, configure the systemby using chaotic particle swarm algorithm, and then insert a record about theconfiguration and job type into the database.