An Artificial Chemistry for Networking

Chemical computing models have been proposed since the 1980ies for expressing concurrent computations in elegant ways for shared memory systems. In this paper we look at the distributed case of network protocol execution for which we developed an online artificial chemistry. In this chemistry, data packets become molecules which can interact with each other, yielding computation networks comparable to biological metabolisms. Using this execution support, we show how to compute an average over arbitrary networking topologies and relate it to traditional forms of implementing load balancing. Our long-term interest lies in the robust implementation, operation and evolution of network protocols, for which artificial chemistries provide a promising basis.