Population dynamic evolutionary algorithm for virtual routers running on multiprocessors system
Xia-an Bi, Qi Sun · 2017
This paper presents an evolutionary game model and proposes a dynamic evolutionary algorithm based on the model. We use the game model with numerous populations to analyze the dynamic selection course of the virtual routers running on multiprocessor systems, and the evolutionary equilibrium is regarded as the solution to this game. Our dynamic evolutionary algorithm is designed with the ability to make the whole system to converge at a stable state. The algorithm is equalizing on an equilibrium fixed point, in which each virtual router tries different processors, and changes the processor selection until a stable proportion distribution is formed. The experimental results show that the dynamic evolutionary approach has very fast convergence speed and is able to effectively balance loads among processors, guarantee fair usage of computing resources among virtual routers and maintain stable high-throughputs of the system.