A methodology for workload characterization of file-sharing peer-to-peer networks

Abstract

The main characteristic of peer-to-peer (P2P) networks is that the hosts in the network may act as both clients and servers at the same time, being called servents. These networks have been widely adopted for sharing idle computational resources available in the Internet, improving content accessibility while reducing costs and response latency, although host availability and content coherence is not usually guaranteed. As a consequence, traditional workload characterization strategies are not suitable for analyzing and understanding these networks, motivating the design of specific strategies for their characterization. In this article we present a novel workload characterization methodology for P2P networks, which account for the main features of these networks. We validate our methodology through the characterization of the Gnutella network, through which we are able to characterize file-sharing patterns, the availability of the servents, and the search patterns, among others.