Peer-to-Peer Network Cost Model
Title: Peer-to-Peer Network Cost Model
Abstract: This research proposes a cost model to evaluate the cost each peer incurs for being part of a peer-to-peer overlay network. Such a cost model has several useful applications, including providing a benchmark to compare between different proposals, predicting disincentives, and designing load balancing primitives. The model considers the load imposed on each node, the distance to other nodes, and the node's degree of connectivity. It is applied to several routing geometries used in recently proposed distributed hash tables (DHT) algorithms. The research concludes with a discussion of open problems this research has uncovered.
Main Research Question: How can we quantify the cost incurred by each peer participating in a peer-to-peer network?
Methodology: The study proposes a cost model that takes into account the load imposed on each node, the distance to other nodes, and the node's degree of connectivity. It assumes a DHT-like structure with vertices (nodes), edges, keys (items), and a hash function that assigns keys to vertices. The model characterizes each request with two independent random variables: the node making the request and the key being requested. It then considers four situations a node may be in when a key is requested: not holding or requesting the key, not on the routing path of the request, holding the key and paying a price for serving the request, requesting the key and paying a price for looking up and retrieving it, and not holding or requesting the key but having to forward the request, thereby paying a price. The overall routing cost experienced by a node is the average over all possible keys of the values of the prices such that the node is on the path of the request.
Results: The study applies the cost model to several routing geometries used in recently proposed DHT algorithms. The results provide insights into the cost incurred by each peer participating in a peer-to-peer network and can be used to compare between different proposals, predict disincentives, and design load balancing primitives.
Implications: The proposed cost model offers a quantitative approach to understanding the cost incurred by each peer participating in a peer-to-peer network. It can be applied to various types of networks, including file-sharing systems, ad-hoc networks, and application-layer multicast overlays. The results can help designers and researchers better understand the trade-offs involved in designing peer-to-peer networks and develop more efficient and cost-effective systems.
Link to Article: https://arxiv.org/abs/0401010v1 Authors: arXiv ID: 0401010v1