An artificial hormone system for self-organization of networked nodes

Abstract

The rising complexity of distributed computer systems give reason to investigate self-organization mechanism to build systems that are self-managing in the sense of Autonomic and Organic Computing. In this paper we propose the Artificial Hormone System (AHS) as a general approach to build self-organizing systems based on networked nodes. The Artificial Hormone System implements a similar information exchange between networked nodes like the human hormone system does between cells. The artificial hormone values are piggy-backed on messages to minimize communication overhead. To show the efficiency of the mechanism even for large scale systems we implemented a simulation environment in Java to evaluate different optimization strategies. The evaluations show that local information is enough to meet global optimization criterion.