Practical Online Network Stack Evolution
Pierre Imai, Christian Tschudin · 2010
Protocol stacks can optimize their performance well for pre-defined environments and are usually hand-engineered in advance: This means that adding new protocols or drastically changing the run-time system or assumed traffic patterns would require re-engineering. As an alternative we present an autonomous stack composition framework inspired by biological evolution: Compositions compete online and new compositions evolve from the best previous performers. We demonstrate the feasibility of our runtime adaptive approach and show through simulations that it can find efficient solutions where offline approaches fail.