On chemical and self-healing networking protocols

In this thesis, we look at networking protocols through the eyes of a chemist. By introducing an artificial chemistry for networking, we obtain an intrinsically dynamic, reaction-based view to packet processing. Our chemical virtual machine lets packets react with each other akin to chemical molecules. Simple reaction rules at the microscopic level translate to well-defined behavior on the macroscopic flow level. Because these two descriptions are linked by laws from chemical kinetics, we are able to apply tools from chemistry to predict the behavior of chemical networking protocols and even proof dynamic convergence properties.
Based on this principle, we develop an engineering model to design and analyze chemical protocols. We demonstrate the feasibility and usefulness of our approach with several new solutions to application scenarios ranging from a gossip-style aggregation protocol, over an enzymatic MAC protocol, to a chemical TCP-like congestion control algorithm, which ensures the coexistence and fairness among chemical and classical packet flows in the Internet.
The chemical reaction model has additional properties that are hard to achieve on traditional code execution platforms: We show how protocol software is able to continuously regenerate itself in order to exhibit intrinsically self-healing properties; our approach is based on self-replicating code and natural selection. We present a self-healing multipath routing protocol that is resilient to the removal of large parts of its own code.
With this work, we try to contribute to the future Internet by discovering the self-regulating capabilities of packet flows, which currently lie dormant.