Economically Scalable Peer-to-Peer Grid Middleware

The usage of economics in grid systems designs has been very popular recently. System designers have attempted to engineer the grid market mechanism that would lead to an efficient outcome before implementing it and open it to end users. It can be noted that, although the usage of economics in grid computing is rather recent, many different grid market mechanisms have been, to this date, designed. Fewer were actually implemented and, in fact, very few, if any, have been successfully used. The argument developed in this article is that this multitude of systems and their relative lack of success is due to the generally adopted ‘top-down’ approach to economic design in grid computing. Such a centralised manner of developing a market structure for a grid environment, although fits well some economic theories and models, is in fact quite the opposite to the way most markets have been created. Instead of the top-down approach to grid market design, this article prescribes a phenomenological approach. It is argued that grid market environments should be built bottomup instead. In this new paradigm, the role of a grid middleware ceases to be prescriptive, in the sense that it carries a (or a set of) particular predefined market mechanism, and becomes a foundation on which the individual choices and interactions of users build an evolving, self-organised, self-adaptive and allocatively efficient grid market. A multi-agent/peer-topeer-based generic middleware is proposed as an illustration of the new paradigm.