A Research On Task Scheduling Algorithm Of Cloud Computing Based On Genetic Algorithm

Cloud computing is a new kind of business computing model. It is connected by network, which make it able to achieve a variety of applications, data and IT services. The core of cloud computing is to manage the resources in the cloud and the tasks that users submitted according to the users’ requirements. And the users only need to pay according to their needs. Thus in the cloud, it is an important issue to consider how to meet the different needs for quality of service (QoS) of different users.Genetic algorithm is a kind of evolutionary algorithm. It is a global optimization search algorithm and based on the thought of biological evolution and the natural selection mechanism of "survival of the fittesf".Genetic algorithm is introduced into the resource scheduling under massive cluster system environment because of its features of parallel and global search in the solution space. This dissertationis based on users’ demand for QoS. By setting the weight vector and considering four factors of different users’ requirements, includingjob completion time, bandwidth, reliability and costs, this dissertation design a fitness function based on users’ degree of satisfaction to ensure the quality of service.Genetic algorithm has shortcomings of "premature". To solve this problem, this dissertation uses simulated annealing algorithm to optimize genetic algorithm. Simulated annealing algorithm is based on the solid annealing mechanism in physics. It has the feature of being able to jump out of local optimum solution. It isa global optimization algorithm.However,it has shortcomings that it does not know much about the entire search space. Thus, combination of genetic algorithm and simulated annealing algorithm can makebest use of both algorithms and bypass their disadvantages,and then improve the performance of both algorithms. This dissertationintroduces simulated annealing operator into genetic algorithm to control the process of generating new individuals. And whethera new individual is accepted or not is decided by Metropolis criterion. In this way, not only the diversity of population can be ensured, but also make the population evolve gradually.This dissertationmakes an introduction to cloud computing simulation tool CloudSim, and introduces the experimental environment configuration.The experiments of the algorithms designed in this dissertation are conducted under the CloudSim simulation platform. Compared with the basic genetic algorithm,the genetic algorithm designed in this dissertationshows better results in meeting the different requirements for cloud QoS of different users. And through comparison among the four algorithms, which are the two algorithmsdesigned in this dissertation, theRandom allocation algorithm (RA) and theRobin-Round scheduling algorithm (RR),the results show thatthe genetic algorithm optimized by simulated annealing operator has a better performance.