PPLive Project

People

Professor Klara Nahrstedt
Professor Indranil Gupta
Jin Liang
Long Vu

 

News

- We are releasing the PPLive crawler (PPCrawLive) (version 1.0). Go the bottom of this page to download. Thank you ! (April 18, 2008)


Project Descriptions

IPTV applications have emerged and opened a new market which is predicted that the number of subscribers increases 10 times to 36.9 million, and the revenues could reach US$10 billion by 2009. This attracts numerous focuses from both academia and industry domains in developing new solutions for this promising market. Among current IPTV technologies, chunk-driven peer to peer (P2P) streaming has been the most successful technology in terms of number of simultaneous viewers. However, a feasible IPTV system for tens folds larger population still remains unanswered. In this project, we concentrate on studying how user preference, content and show-time of program influence user behaviors and system overlay. To our knowledge, this relationship plays a crucial role in large scale P2P media streaming systems.

Several deployed p2p file sharing overlays have been characterized by recent studies in the literature. In this project, we show that when one considers a p2p application that streams media instead of sharing files, many conclusions of these previous studies become false. Specifically, we undertake a crawler-based investigation of PPLive, the largest live multimedia streaming system in the world today. Another motivation for our work comes from the recent growth in popularity of IPTV - understanding overlays like PPLive will be critical to building larger-scale media streaming overlays in the future. Our task is made challenging by the fact that PPLive is proprietary and very few internal design decisions are known. PPLive has multiple channels, and each channel has its own overlay. Each channel streams either live audio-video feeds, or movies according to a preset schedule. A user that is subscribed to any channel joins the PPLive network, but the user's client machine may be used to relay feeds for channels other than the subscribed one. Popular PPLive channels often contains several tens of thousands of nodes.

Our experiments were done by crawling the real running PPLive network. Our crawlers were run on both machines in the CSIL (Computer Science Instruction Laboratory) cluster at UIUC, as well as on 10 PlanetLab hosts. Our major findings are that: (1) Unlike p2p file sharing users, PPLive peers are impatient, (2) Channel Size variations are higher than in p2p file sharing networks, (3) Average degree of a PPLive peer is independent of the channel size, (4) Small PPLive overlays (channels) are more similar to random graphs in structure, compared to large PPLive overlays, (5) PPLive peer pairs have a bimodal distribution in their availability correlation, i.e., their availabilities are either highly correlated or not at all. We believe these results caution us against hastily reaching generic conclusions about overlay characteristics, and point us towards taking seriously the nature of applications while designing and optimizing p2p overlays.

Publications

1.Understanding Overlay Characteristics of a Large-scale P2P IPTV System, Long Vu, Indranil Gupta, Klara Nahrstedt, Jin Liang. ACM Transactions on Multimedia Computing, Communications and Applications (TOMCCAP), Vol. 6, No. 4, 2010.
2. Understanding Overlay Characteristics of a Large-scale P2P IPTV System, Long Vu, Indranil Gupta, Klara Nahrstedt, Jin Liang. UIUC Tech report (09/2008)
3. Measurement and Modeling a Large-scale Overlay for Multimedia Streaming, Long Vu, Indranil Gupta, Jin Liang, Klara Nahrstedt. International Conference on Heterogeneous Networking for Quality, Reliability, Security and Robustness (QShine 2007)
4. Measurement of a Large-scale Overlay for Multimedia Streaming, Long Vu, Indranil Gupta, Jin Liang, Klara Nahrstedt. High Performance Distributed Computing (Poster - HPDC 2007)
5. Mapping the PPLive Network: Studying the Impacts of Media Streaming on P2P Overlays, Long Vu, Indranil Gupta, Jin Liang, Klara Nahrstedt. UIUC Tech report (UIUCDCS-R-2006-275), 08/2006

Funding Agencies

The PPLive project is supported in part by NSF CAREER grant CNS-0448246, NSF ITR grant CMS-0427089, NSF ANI 03-23434, and NSF CNS05-09314. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation or US government.

Sharing the Trace

We share the real trace we obtained from our experiments with PPLive. There are two main traces: snapshot and partner list. The former is obtained by Snapshot Operation and the later is obtained by Partner Discovery Operation. Notice that you will take your own risk to use these traces. There is no guarantee about the correctness of the traces because (1) PPLive is a closed-source application, and (2) many PPLive clients are inside NAT and firewalls and thus we can not probe them. You can use the trace of Partner Discovery to construct a (incomplete) topology of PPLive.
Snapshot Trace
Partner Discovery Trace

Please read the README.txt files carefully before you dig in the traces. Feel free to email me if you have questions. Please acknowledge if you publish any paper using these traces. Thank you!

Sharing the PPLive crawler code (PPCrawLive)

As many researchers in Brazil, France, Korea, China, USA and Taiwan email and request for the crawler code. We decide to release the code. Click this link to download the code: PPCrawLive

Please read the README.pdf files carefully before you start. Feel free to email me if you have questions. Please acknowledge if you publish any paper using this crawler. Thank you!

 


Contact

Direct comments and suggestions to Long Vu (longvu2@uiuc.edu)
Copyright © 1997-2007, Multimedia Operating Systems and Networking Group, University of Illinois at Urbana-Champaign, IL61801.