An architecture for internet-based distributed evolutionary computation

Abstract

In this paper, an Internet-friendly architecture for distributed evolutionary computation is introduced. To make the architecture more flexible and fault-tolerable, it defines server (master) as a more important role. It serves not only as a simple communication center that transmits individuals during the migration operation, but also to maintain sub-populations. Also the progress of evolution is discredited into small tasks to assign to the clients (slaves). By working this way, it is also able to allocate computing resources dynamically, and is more efficient especially in heterogeneous environments. Browser/Server model is used in it. Native Client is used in Browser as the container to provide good performance for algorithms. © 2012 Springer-Verlag.