ABSTRACT
The effective use of distributed computation is a challenging task, since the processes must obtain resources in a dynamically changing environment and must be designed to collaborate despite a variety of asynchronous and unpredictable changes. From a scientific point of view, the analogy between distributed computation and natural ecologies brings to mind the spontaneous appearance of organized behaviour in biological and social systems, where agents can engage in cooperating strategies while working on the solution of particular problems. The chapter describes a market-based computational ecosystem and a theory of distributed computation. The theory describes the collective dynamics of computational agents, while incorporating many of the features endemic to such systems, including distributed control, asynchrony, resource contention, and extensive communication among agents. The chapter provides an effective procedure for controlling chaos in distributed systems. It is based on a mechanism that rewards agents according to their actual performance.
