An Electronic Marketplace: Agent-based Coordination Models for Online Auctions

Different coordination models exist for managing concurrency in complex distributed systems, for example, direct interaction, tuple spaces, hierarchical structures, and publish-and-subscribe mechanisms. Online auctions are complex distributed systems, where the choice of coordination model can significantly impact the performance of multiple software agents, acting on behalf of human buyers and sellers. In this paper, we evaluate analytically and experimentally different auction types and coordination models in an electronic marketplace. Three important metrics are: the number of software agents, the number of messages exchanged between these agents, and the number of migrations performed by the agents. We conclude that dynamically choosing the number of agents and the coordination model can improve quality of service in electronic commerce applications. Furthermore, in online auctions where there is a lot of interaction between software agents, migration of agents to virtual stores and local bidding is more fair and efficient than using remote interaction. This observation has the potential to significantly improve and redefine existing online auction systems such as eBay.