A novel superior peer selecting algorithm in large-scale DHT systems

Flexible superior peer selecting techniques are of particular interest for the study of scalability of the P2P system, especially in many structured P2P systems, such as KAD, Chord, and CAN etc., since most of these systems need to change flat topology to hierarchical topology to reduce lookup time delay and maintenances overhead as more and more users join into them, here the fundamental challenge is to identify and select these types of super nodes to achieve system scalability with minimum disruption to original protocols. However, a few efforts have been made until now to use survival analysis theory, which is the most common methods in biomedicine, to resolve the superior peer selecting problem. In the attempt to fill this gap, several survival analysis methods are introduced in this paper. The content of our research includes exploring the critical factors that may influence longevity of peers and building recurrent events based lifetime model using counting process. The study of this model is aimed to analyze and predict remaining uptime of a peer in a large scale Distributed Hashing Table (DHT) system. Based on these work, a simple yet powerful superior peer selecting mechanism is provided. The simulations prove that our new algorithm is reliable and effective.