PPLive Project
People
Professor Klara Nahrstedt
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.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.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)