The optimization of replica distribution in the unstructured overlays

Abstract

Replication is a widely used technique in unstructured overlays to improve content availability or system performance. A fundamental question often addressed by previous work focused on: how many replicas ought to be allocated for each data item given the fixed query rates and limited storage capability? In this paper, we have put forth two optimal replica distributions to achieve the highest success rate and the lowest message consumption. Especially, we have investigated the influence of item size on replica distribution. Our results show that Square-Root Replication, which is traditionally considered to be optimal, is not always the best choice. Our study offers a new deep understanding of resource managment in self-organized unstructured overlays.