Evaluation Of P2p Systems Under Different Churn Models: Why We Should Bother

Research on peer-to-peer (P2P) systems has been hampered by the fact that few systems are actually in use, and the space of possible applications is still under scrutiny. As a consequence, new ideas have been mostly evaluated using synthetic data, traces from a few existing systems and simulators, with a poor characterization of churn. This void has lead to the formulation of a variety of models, with implications that have not yet been made altogether clear to the community. In this work, we discuss the question whether it pays off to evaluate P2P applications using more than one churn model. Although an affirmative response could appear to be obvious at first glance, we show that depending on the aspects under consideration, models can yield equivalent results, saving implementation time but leading to spurious generalizations if proper care is not taken.