ABSTRACT

Distributed Hash Tables (DHTs) are an important tool in building wide-area peer-to-peer applications. To achieve scalable performance, a node in a DHT system maintains links with a subset of all nodes in the overlay network. These links can have a significant effect on the overall performance of the DHT system. The network proximity of the links stored in the DHT will affect performance when used to search for objects. In this paper, we evaluate the performance of joins in Chord [10] and present some modified join algorithms. Our results indicate that with some modest increases in insert costs and a decrease in loadbalancing, our method can improve lookup latency by thirty percent.