Duration: 1:19:58

20 views

0

During recent years, the pervasive deployment of Peer-to-Peer (P2P) systems had a profound impact on the Internet that is even more tangible than the impact of the Web. Ease of deployment and self-scaling are two key factors that continue to fuel the growing popularity of the P2P communication paradigm for a wide spectrum of large scale commercial systems ranging from content distribution to Internet telephony (e.g. Skype) and Internet TV. Despite the importance of the P2P communication paradigm and the far reaching impact of P2P systems on the Internet, fundamental properties and in particular the dynamics of large scale P2P systems are not well understood. In this talk, I present an overview of the IonP2P project. The goal of this project is to investigate and develop new measurement and modeling methodologies to  understand and accurately characterize properties and dynamics of large scale P2P systems. First I show that the key challenge in accurately characterizing large scale P2P systems is to capture their representative (i.e. unbiased) snapshots. I demonstrate how the dynamic and heterogeneous nature of real world P2P systems can introduce bias into the selection of representative samples of peer properties (e.g. degree, link bandwidth, number of files shared). Then, I propose the Metropolized Random Walk with Backtracking (MRWB) as a viable and promising technique for collecting nearly unbiased samples. Our extensive simulation study and empirical evaluation demonstrate that MRWB works well for a wide variety of commonly-encountered P2P network conditions. Time permitting, I will also present an overview of PRIME, our mesh-based P2P streaming mechanism for live video. Further information on the IonP2P project is available at (http://mirage.cs.uoregon.edu/P2P/). The IonP2P project is funded in part by NSF and Cisco.