Generalized Distance Metric for Various DHT Routing Algorithms in Peer-to-Peer Networks

TL;DR

This work addresses unifying diverse DHT routing schemes (Chord, Kademlia, Tapestry, Pastry) under a single generalized distance metric. It defines $D = \sum_{i=0}^{k-1} \left\{\left(r_i-h_i+2^d\right) \\bmod 2^d\right\} \times 2^{d \times i}$ over digitized node IDs and hash IDs, showing that varying the digit size $d$ and length $k$ reproduces the four algorithms, including special-case forms for Chord/Pastry, Kademlia, and Tapestry. The paper further argues that routing tables may vary in size across nodes without changing the root node for a given key, as long as the shared metric is used, and it demonstrates routing behavior with concrete examples. It highlights the practical implications for memory-hop trade-offs in large DHTs, offering a flexible framework for routing design in resource-constrained peer-to-peer networks.

Abstract

We present a generalized distance metric that can be used to implement routing strategies and identify routing table entries to reach the root node for a given key, in a DHT (Distributed Hash Table) network based on either Chord, Kademlia, Tapestry, or Pastry. The generalization shows that all the above four DHT algorithms are in fact, the same algorithm but with different parameters in distance representation. We also proposes that nodes can have routing tables of varying sizes based on their memory capabilities but with the fact that each node must have at least two entries, one for the node closest from it, and the other for the node from whom it is closest in each ring components for all the algorithms. Messages will always reach the correct root nodes by following the above rule. We also further observe that in any network, if the distance metric to define the root node in the DHT is same at all the nodes, then the root node for a key will also be the same, irrespective of the size of the routing table at different nodes.