Hypernets: Good News for Gnutella

Title: Hypernets: Good News for Gnutella

Abstract: The Gnutella network, a type of peer-to-peer (P2P) network, has been criticized for its scalability, particularly due to its original tree-like structure. This structure, known as a Cayley tree, is known to have lower bandwidth than higher-dimensional topologies. However, the Gnutella network is virtual and can use hyper-topologies, which are not limited by hardware constraints. This study compares the query throughput of several hyper-topologies, including virtual hypercubes and virtual hypertori, up to millions of peers. The results show that these hyper-topologies offer near-linear scalability, making them promising solutions for improving the scalability of the Gnutella network.

Main Research Question: Can hyper-topologies improve the scalability of the Gnutella network?

Methodology: The study uses mathematical models to compare the query throughput of several hyper-topologies in the Gnutella network. The models assume that the distribution of peer activity is similar to that used by Ritter (2001). The study focuses on the virtual hypercube and the virtual hypertorus, which are higher-dimensional topologies that can potentially improve the network's scalability.

Results: The study finds that the virtual hypercube and the virtual hypertorus offer near-linear scalability up to millions of peers. This means that the bandwidth of these hyper-topologies increases linearly as the number of peers increases, which is a significant improvement over the original Cayley tree structure.

Implications: The results of this study suggest that hyper-topologies can significantly improve the scalability of the Gnutella network. By using higher-dimensional topologies, the network can support a larger number of simultaneous peers without experiencing a decrease in bandwidth. This could lead to faster query times and a more efficient use of resources, making the Gnutella network a more viable option for large-scale P2P applications.

Link to Article: https://arxiv.org/abs/0202019v1 Authors: arXiv ID: 0202019v1