Modeling and performance analysis of BitTorrent-like peer-to-peer networks
Proceedings of the 2004 conference on Applications, technologies, architectures, and protocols for computer communications
SIGMETRICS '05 Proceedings of the 2005 ACM SIGMETRICS international conference on Measurement and modeling of computer systems
Building scalable and robust peer-to-peer overlay networks for broadcasting using network coding
Proceedings of the twenty-fourth annual ACM symposium on Principles of distributed computing
Incentives in BitTorrent induce free riding
Proceedings of the 2005 ACM SIGCOMM workshop on Economics of peer-to-peer systems
Analyzing BitTorrent and related peer-to-peer networks
SODA '06 Proceedings of the seventeenth annual ACM-SIAM symposium on Discrete algorithm
MTcast: robust and efficient p2p-based video delivery for heterogeneous users
OPODIS'05 Proceedings of the 9th international conference on Principles of Distributed Systems
Cost minimisation in multi-interface networks
NET-COOP'07 Proceedings of the 1st EuroFGI international conference on Network control and optimization
Cost minimisation in unbounded multi-interface networks
PPAM'07 Proceedings of the 7th international conference on Parallel processing and applied mathematics
| Hi-index | 0.00 |
In this paper we analyze the ability of peer to peer networks to deliver a complete file among the peers. Early on we motivate a broad generalization of network behavior organizing it into one of two successive phases. According to this view the network has two main states: first centralized – few sources (roots) hold the complete file, and next distributed – peers hold some parts (chunks) of the file such that the entire network has the whole file, but no individual has it. In the distributed state we study two scenarios, first, when the peers are “patient”, i.e, do not leave the system until they obtain the complete file; second, peers are “impatient” and almost always leave the network before obtaining the complete file. We first analyze the transition from a centralized system to a distributed one. We describe the necessary and sufficient conditions that allow this vital transition. The second scenario occurs when the network is already in the distributed state. We provide an estimate for the survival time of the network in this state, i.e., the time in which the network is able to provide all the chunks composing the file. We first assume that peers are patient and we show that if the number of chunks is much less than en, where n is the number of peers in the network, then the expected survival time of the network is exponential in the number of peers. Moreover we show that if the number of chunks is greater than $\frac{\log n}{n+1}e^{n+1}$, the network's survival time is constant. This surprisingly suggests that peer to peer networks are able to sustain only a limited amount of information. We also analyze the scenario where peers are impatient and almost always leave the network before obtaining the complete file. We calculate the steady state of the network under this condition. Finally a simple model for evaluating peer to peer networks is presented.