Gnutella represents a popular class of peer-to-peer (P2P) networks that are known as unstructured P2P networks, and has served as a real-world testbed for measurement-based characterization of these networks. Due to its open nature, the Gnutella protocol has undergone a series of gradual improvements over the years. To address the well known scalability problem with flood-based search techniques in the face of growing user population, two key features were introduced to the protocol: (i) a semi-structured topology, and (ii) a dynamic querying mechanism. Despite its importance, to our knowledge, no characterization of the Gnutella topology has been conducted during the last few years.

In this paper, we present a detailed characterization of the Gnutella overlay topology based on a recent measurements. We present a set of techniques (i) to efficiently capture accurate snapshots of the Gnutella network, and, more importantly, (ii) to properly quantify the accuracy of the captured snapshots. Using a new crawler that incorporates these techniques, we characterize different properties of today's Gnutella topology, examine their underlying causes, and investigate their implications. Our characterizations not only shed light on the current status of the Gnutella network but more importantly provide a better understanding of several fundamental challenges in the design of unstructured overlays.