On the distribution of nodes in distributed hash tables

Abstract

We develop a model for the distribution of nodes in ring-based DHTs like Chord that position nodes randomly or based on hash-functions. As benefit of our model we get the distribution of interval sizes and an approximation of the distribution of data load among the participating peers. The results match previously published experimental studies on load distribution that motivated the research on load balancing. Most importantly, our model also gives a theoretical explanation for observed load behavior.