Towards Self-Organized Load Distribution over Chaotic Resources

Abstract

This paper addresses a question of whether resources suffering nonlinear fluctuations can maintain their stability as a system expands for computing tasks in a distributed manner. To this end, we suggest that by evolving individual resources following the self-organized criticality of sandpile model, the whole load distribution system can reach a stable state after a small but extremely local overhead occurs, leading to lots of avalanches. The proposed load balancing approach is evaluated in terms of latency minimization.