A Distributed Game-theoretic Approach for IaaS Service Trading in an Auction-based Cloud Market

摘要

With the rapid development of IaaS market, how to efficiently trade between cloud providers and users is becoming a new challenge which attracts huge attentions from both industry and academia. Compared with traditional fixed-price model, market-oriented trading mechanism such as auction demonstrates greater promise for resource pricing and allocation in clouds due to its adaptability and flexibility. In this paper, we focus on the competitive and bidding scenario among independent users in an auction-based IaaS market. Participants have different composite service demands and they adjust their bidding prices during the trading with the aim of obtaining appropriate amount of resources to maximize their profits. We formulate this scenario as a dynamic noncooperative game with incomplete information and propose a feedback-based distributed bidding adjustment approach to find the approximated optimal bids (i.e. Nash Equilibrium) for each user so as to achieve a fair and multi-win resources allocation outcome in the whole market. Our experimental investigation showed that with the help of our agent-based automated trading model and distributed bidding algorithms, the Nash equilibrium state of the cloud resource allocation game can be reached efficiently within an acceptable time, and the optimal bidding prices of each user could be obtained at the same time.