An Auction and League Championship Algorithm Based Resource Allocation Mechanism for Distributed Cloud

Abstract

In cloud computing, all kinds of idle resources can be pooled to establish a resource pool, and different kinds of resources combined as a service is provided to users through virtualization. Therefore, an effective mechanism is necessary for managing and allocating the resources. In this paper, we propose a double combinatorial auction based allocation mechanism based on the characteristics of cloud resources and inspired by the flexibility and effectiveness of microeconomic methods. The feedback evaluation based reputation system with attenuation coefficient of time and the hierarchy of users introduced is implemented to avoid malicious behavior. In order to make decisions scientifically, we propose a price decision mechanism based on a BP (back propagation) neural network, in which various factors are taken into account, so the bidding/asking prices can adapt to the changing supply-demand relation in the market. Since the winner determination is an NP hard problem, a league championship algorithm is introduced to achieve optimal allocation with the optimization goals being market surplus and total reputation. We also conduct empirical studies to demonstrate the feasibility and effectiveness of the proposed mechanism.