Chemotaxis-inspired load balancing

Abstract

We present an approach to the problem of load balancing on networks of nodes. Our approach is inspired by the phenomenon of negative chemotaxis in living systems. We use a diffusing signal (which is emitted by load, and moves faster than the load) to guide the movement of load towards the balanced state. Our reference system (for comparison) is unguided, diffusing load, moving at the same speed. Our tests show that the chemotaxis system can give large improvements over the reference system in convergence speed, as well as showing much reduced sensitivity to variations in network topology and in initial load distribution. Copyright © 2006 S. Karger AG.