Research On QoS-driven Resource Allocation Method In Cloud Computing Environment

Nowadays, cloud computing has become a booming business compute model, it is the most widely studied and applied computing method among the distributed computing academic sphere. Cloud computing include a large number of servers and numerous users which means frequent resource allocation is needed. At present research on how to design an efficient resource allocation method which based on the satisfy user's different demands for resource has become a hotspot in cloud computing sphere.The existing cloud computing resource allocation did not consider the feature of the service-oriented, the resource allocation objective is simple and can not meet users' QoS(Quality of Service) demand. With the continuous development of cloud computing, it's scale is becoming larger and larger, so the allocation of resources is becoming more and more complex. Using single heuristic algorithm has certain limitation, it is hard to improve the performance and convergence of the algorithm. In the view of the above problem, this dissertation proposed a resource allocation model which has double objective constraint, and designed a joint optimization algorithm JOA(Joint Optimization Algorithm, JOA) to solve the resource allocation in cloud computing environment. This dissertation has mainly done these works:(1) Proposed a resource allocation model with double objective constraint. On the basis of the business model of cloud computing, proposed the objects of resource allocation, taking both the different requirement of user task's deadline and the cost minimization into account, finally established the mathematical model for the resource allocation.(2) To solve the problem of QoS-Driven resource allocation in cloud computing environment, a joint optimization algorithm named JOA has been designed. The joint optimization algorithm JOA, first by introducing penalty factor take the two scheduling objectives into a single objective function, then design the fitness function and propose adaptive adjusting crossover probability. Dynamically switch from genetic algorithm to ant colony algorithm use the strategy designed by evolutionary rates of genetic algorithm. Using genetic algorithms initialize the pheromone of ant colony algorithm by a pheromone transformation strategy. Using the current state of the virtual machine nodes to calculate the heuristic information, and introduce the stupid ants in ant colony algorithm implementation process to increase the random selection of a virtual machine.(3) Analyzed the simulation experimental results. Finally, this dissertation has implemented the resource allocation method on the CloudSim simulation platform. The results have proved the joint optimization algorithm JOA is more effective than genetic algorithm and ant colony algorithm in solving QoS-driven resource allocation.