A market mechanism for participatory global query: A first step of enterprise resources self-allocation

• Authors: Cheng Hsu;Christopher D. Carothers;David M. Levermore
• Venue: Information Technology and Management
• Year: 2006

The problem of Database Query has always been considered from the user's side. That is, the databases are always treated merely as the object of search, rather than being a subject or willing participants of an information exchange. This paradigm works when all participating databases belong to a single authority (such as a company) under which their participation is definitive and their contents completely open for the querying. Traditional single databases, federated databases, and even the new XML-based Internet databases subscribe to this user-oriented paradigm. However, emerging information enterprises are increasingly collaborative in nature, since they tend to involve, on a real-time and on-demand basis, a large number of databases belonging to many different organizations whose participation is conditional and case-by-case; e.g., drilling through supply chains. These collaborative queries deserve a new paradigm that equally account for the provider side. Research has shown that market-style self-allocation of users to providers is a promising approach to support such a paradigm. However, previous results of artificial markets are insufficient for global database query. Therefore, we develop an artificial market model to provide a Two-Stage Collaboration solution, where the first stage establishes optimal participation of databases for a search task, and the second executes the task in a traditional database query manner. The proposed model employs a new agent-based, peer-to-peer publish and subscribe approach to self-allocating database resources in an information enterprise. This approach promises to lead eventually to allocating other classes of information resources, as well. New results include (1) an agent model using a Metadatabase and an Agent-Base to create and manage large number of custom agents, (2) a peer-to-peer negotiation method, and (3) an open common schema design. The paper also provides an implementation scheme for developing the artificial market. Laboratory tests show that such a mechanism is feasible for large scale matching and negotiation as required by the first stage. The second stage employs mainly previous results established in the field.