Static Resiliency vs Churn-Resistance Capability of DHT-Protocols

During the last few years, the topic of peer-to-peer networking (P2P hereafter) has gained significant attention from the research community. The focus has mainly been on the design, implementation and evaluation of systems based on Distributed Hash Tables (DHTs), as these structures have potential to become a general purpose substrate for distributed applications. Nonetheless, there are still dimensions of this technology that just have been barely studied and require further investigation before the full potential of P2P technologies could be achieved. A specific aspect of this networking technology requiring attention relates to the dynamic nature of P2P overlay networks. The composition of overlay networks is typically highly dynamic and heterogeneous in multiple ways, thereby making difficult the task of providing guarantees for the performance metrics of the system. More specifically, the continuous arrival and departure of new (or returning) nodes, commonly referred to as churn, is a problem that has not been extensively studied, and remains a major problem to be tackled. A particular aspect of churn is its impact on the overall performance of P2P networks. This paper addresses the impact of churn on P2P networking performance and presents a framework to evaluate different aspects of P2P overlay networks. The focus of this paper is on structured, DHT-based overlay networks. The main contributions of this paper are a new insight into the impact of churn on P2P overlay network performance, a set of new evaluation criteria that captures the impact of churn on P2P performance, and a detailed set of experiments which uses these evaluation criteria to analyze the performance of multiple DHT architectures under churn.