This paper builds a complete modeling framework for understanding user churn and in-degree dynamics in unstructured P2P systems in which each user can be viewed as a stationary alternating renewal process. While the classical Poisson result on the superposition of n stationary renewal processes for n requires that each point process become sparser as n increases, it is often difficult to rigorously show this condition in practice. In this paper, we first prove that despite user heterogeneity and non-Poisson arrival dynamics, a superposition of edge-arrival processes to a live user under uniform selection converges to a Poisson process when system size becomes sufficiently large. Using this finding, we then obtain closed-form results on the transient behavior of in-degree, paving novel ways for a variety of additional analysis of decentralized P2P systems. Categories and Subject Descriptors C.4 [Performance of Systems]: Modeling techniques General Terms Algorithms, Performance, Theor...
Zhongmei Yao, Daren B. H. Cline, Dmitri Loguinov