Design And Simulated Implementation Of The Auction Mechanism For Grid Resources

Various resources are interconnected and aggregated in grid, and the spare resources can be provided to the users who need them best. In order to motivate providers to actively provide their own idle resources, the grid needs to establish an allocation mechanism which can both encourage providers greatly and take fully into account the interests of both users and providers.As a result, a lot of resource scheduling models based on the economy principles are proposed. Some of them are price-based models in which the providers set prices. The others are negotiation-based models in which price is done by both providers and the users. As one of the negotiation models, auction can provide current supply and demand information, and can help the users and the providers evaluate resources. In this thesis, resources are classified into general resources and special resources. Based on the various users’requirements differently, three auction mechanisms are designed to trade short-term and long-term resources’spare time. When short-term general resources are traded, the unit price can be adjusted according to supply and demand information and a proper scheduling method will be choosen to allocate the users’required resources. When long-term general resources are traded, several providers’resources are bidded in an auction simultaneously. Every bidder can have more choices and the profits among the suppliers can be balanced. When special resources are traded, a proper mechanism is designed to make it fit for users to rent some period of resources. Genetic algorithm is used to solve the combinatorial optimization problems encountered in the three mechanisms. According to the results, resources are scheduled.In order to check the feasibility and the validity of the designed mechanisms, simulated experiments are made to build the simulated models based on the auction mechanisms using Java language with Simjava toolkit, and the analysis of the performance of the related algorithm have been carried on. Simulation results have shown that the mechanisms are both feasible and effective.