Research On Resource Allocation Strategy Based On Market Auction Mechanism In Cloud Computing

Cloud computing is a new commercial computing model which integrates many related technologies such as distributed computing, utility computing, grid computing and so on. Using developed virtualization technology, cloud computing abstracts large-scale homogeneous or heterogeneous resources as a service which is provided to cloud users on demand via the Internet. Research on resource allocation has been a focal point in current cloud computing, and its commercial and economic characteristics have promoted many researchers to utilize economic theories and methods to solve resource allocation which have achieved some results. On the basis of analyzing the present research situation of resource allocation in cloud computing by using the theories and methods of economics, the auction model, which is an important branch of economics, is introduced into the cloud computing resource allocation field in this thesis. The main contents are as follows:Combined with the characteristics of combinatorial auction model and double auction model, a resource allocation strategy in Cloud computing based on combinatorial double auction model which improves and perfects the traditional auction model is proposed in this thesis. The model can set up the reasonable pricing strategies for provider and user from different angles: the resource node credibility and the task response time are mainly taken into consideration in user’s bidding strategy and the current system load is taken into account in provider’s asking price strategy. Besides, in order to carry out the resource auction multiple rounds to improve resource utilization and user satisfaction, the model also designs price adjustment strategies. Meanwhile, breach penalty mechanism is introduced in the thesis for the purposes of standardizing auction activities. When service level agreement signed between user and provider is not met, provider needs to provide some compensation for user according to the conventions of the agreement.In view of this situation that most cloud resource allocation strategies at present based on auction mechanism consider only monetary benefits but neglect the environmental benefits, this thesis presents a resource allocation strategy considering energy consumption factor. With the improvement of Cobb Douglas production function model, a utility function which takes both the service provider revenue and energy cost into account for provider is designed.Through the expansion of cloud simulation platform CloudSim, two resource allocation algorithms proposed in this thesis are programmed. Experimental results verify that the algorithms proposed are reasonable and effective. Moreover, the simulation analyzes performance of the algorithms proposed in this thesis, shortest job first and CloudSim respectively from the task execution success rate, SLA’s breach, the provider benefits and utility. The results show that the algorithms proposed in this thesis have more advantages in above indicators.