Multi‐factors oriented study of P2P Churn
Author(s) -
Yang Dong,
Zhang Yuxiang,
Zhang Hongke,
Wu TinYu,
Chao HanChieh
Publication year - 2009
Publication title -
international journal of communication systems
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.344
H-Index - 49
eISSN - 1099-1131
pISSN - 1074-5351
DOI - 10.1002/dac.1001
Subject(s) - computer science , causality (physics) , protocol (science) , successor cardinal , data science , data mining , medicine , mathematical analysis , physics , alternative medicine , mathematics , pathology , quantum mechanics
Abstract The dynamics of peers, namely Churn, is an inherent property of peer‐to‐peer (P2P) systems and is critical to their design and evaluation. Although every excellent P2P protocol has some solution to this issue, studies on Churn are still seldom. This paper studies various factors related to Churn, and uses them to analyze and evaluate P2P protocols. Prior researches on Churn are all based on the P2P network factors in Churn environment, and their difference is whether to use these factors as predecessor references to build Churn analytical models or as successor references to test the models. According to this difference, this paper first divides various factors into two categories: impacting Churn and affected by Churn. There is a causal relationship between these two categories. Factors impacting Churn are cause, and the factors affected by Churn are effect. In this paper, we use this causality to simulate and analyze P2P Churn. Cause is used as the input data and effect is used as the output result. Second, based on the classification of Churn factors, we present a performance evaluation framework and two comparing models. Based on the framework and models, we simulate and analyze three P2P protocols and get some useful results such as the performance of these protocols under Churn, the advantage of Chord over others, and the most important factors impacting Churn. Finally, we present a method to improve recent P2P Churn models by adding some influence factors. Copyright © 2009 John Wiley & Sons, Ltd.