Understanding churn in peer-to-peer networks

Abstract

The dynamics of peer participation, or churn, are an inherent property of Peer-to-Peer (P2P) systems and critical for design and evaluation. Accurately characterizing churn requires precise and unbiased information about the arrival and departure of peers, which is challenging to acquire. Prior studies show that peer participation is highly dynamic but with conflicting characteristics. Therefore, churn remains poorly understood, despite its significance.In this paper, we identify several common pitfalls that lead to measurement error. We carefully address these difficulties and present a detailed study using three widelydeployed P2P systems: an unstructured file-sharing system (Gnutella), a content-distribution system (BitTorrent), and a Distributed Hash Table (Kad). Our analysis reveals several properties of churn: (i) overall dynamics are surprisingly similar across different systems, (ii) session lengths are not exponential, (iii) a large portion of active peers are highly stable while the remaining peers turn over quickly, and (iv) peer session lengths across consecutive appearances are correlated. In summary, this paper advances our understanding of churn by improving accuracy, comparing different P2P file sharing/distribution systems, and exploring new aspects of churn.

Deep Inspection of Unreachable BitTorrent Swarms

“…Especially,we evaluate the churn [11] and peer's degree distribution. We use four measurement methods to measure the target swarm.…”

“…To understand the churn characteristics, the session length is the most important property [11]. The session length indicates how long peers remain in the network.…”

“…Last, to evaluate the impact of different measurement methods on peer-to-peer (P2P) characterization, we compare the characteristics of the target P2P network measured concurrently using different methods, such as churn [11] and peer's degree distribution. We show that the results exhibit significant differences according to the measurement methods, and that our method captures the most detailed information of all, revealing that the other methods may result in false characterization of P2P networks due to a lack of detail.…”

Semantics‐Based Resource Discovery in Large‐Scale Grids

“…Even worse, the overwhelming maintenance overhead does not bring much benefit in this situation, because the newly constructed topology will change quickly. Luckily, churn of the nature described above rarely happens in reality [31], and we can see from Figure 12 that with this churn rate, ShortCut degrades to random-walk. The high overhead problem of OntoSum in very dynamic environments can be solved by a simple solution: when the network is very dynamic, the system can give up the ontology-based topology construction and routing and resort to basic Gnutella random-walk as the solution.…”

Decentralized Polling with Respectable Participants

“…To put this duration into perspective, several trace-driven analysis of churn in various peer-to-peer systems and applications, including file-sharing (BitTorrent, eDonkey [34]), Voice over IP [37] and on-line social networks (Facebook, HiFive [30]), report an average session time of at least a few minutes. They further show that this duration is significantly increased when considering peers who have already spent some time in the system, and initial durations of 5 min are typical for social networks.…”