“Peer-to-peer” (or P2P) refers to a class of systems and applications that employ distributed resources to perform a critical function in a decentralized manner. The resources encompass computing power, data (storage and content), network bandwidth, and presence (computers, human, and other resources). Among the critical functions that have to be performed by P2P systems, we can cite distributed computing, data/content sharing, communication and collaboration, or platform services.
Published in Chapter:
Peer-to-Peer Usage Analysis
Florent Masseglia (INRIA Sophia Antipolis, France), Pascal Poncelet (EMA-LGI2P Site EERIE, France), and Maguelonne Teisseire (LIRMM UMR CNRS 5506, France)
Copyright: © 2009
|Pages: 6
DOI: 10.4018/978-1-60566-014-1.ch154
Abstract
With the huge number of information sources available on the Internet and the high dynamics of their data, peer-to-peer (P2P) systems propose a communication model in which each party has the same capabilities and can initiate a communication session. These networks allow a group of computer users with the same networking program to connect with each other and directly access resources from one another. P2P architectures also provide a good infrastructure for data and computer intensive operations such as data mining. In this article we consider a new data mining approach for improving resource searching in a dynamic and distributed database such as an unstructured P2P system, that is, in Masseglia, Poncelet, and Teisseire (2006) we call this problem P2P usage analysis. More precisely we aim at discovering frequent behaviors among users of such a system. We will focus on the sequential order between actions performed on each node (requests or downloads) and show how this order has to be taken into account for extracting useful knowledge. For instance, it may be discovered, in a P2P file sharing network that for 77% of nodes from which a request is sent for “Mandriva Linux,” the file “Mandriva Linux 2005 CD1 i585-Limited- Edition-Mini.iso” is chosen and downloaded; then a new request is performed with the possible name of the remaining iso images (i.e., “Mandriva Linux 2005 Limited Edition”), and in the large number of returned results the image corresponding to “Mandriva Linux 2005 CD2 i585-Limited-Edition-Mini.iso” is chosen and downloaded. Such knowledge is very useful for proposing the user with often downloaded or requested files according to a majority of behaviors. It could also be useful in order to avoid extra bandwidth consumption, which is the main cost of P2P queries (Ng, Chu, Rao, Sripanidkulchai, & Zhang, 2003).