P2P Applications - Peer-to-Peer Computing: Applications, Architecture, Protocols, and Challenges - page 21

that such chunks will not be “extinct” when the playback times come.
Doing this will also enhance the overall availability of packets. In the
anchor-based strategy, the peer tries to get all the chunks located at pre-
defined anchor points used for supporting fast-forward and fast-rewind
functions.
Transmission Strategies. The most important goals of transmission of
video chunks are to maximize download rate and to minimize overheads.
However, these goals are often times conflicting with each other. The rea-
son is that a commonly used e cient way to maximize download rate
is to request chunks from multiple neighboring peers at the same time.
While such parallel downloading can speed up the transmission process,
inevitably some chunks are received in duplicates, thereby leading to
overheads. Indeed, in practical video streaming systems, a downloading
peer usually employs one or more of the following three strategies:
•Downloading the same chunk from multiple peers at the same time;
•Downloading several different chunks from multiple peers at the
same time; and
•Downloading all chunks from one selected peer and switch to an-
other peer if the former is unavailable.
In the following we go through a brief survey of representative P2P video
streaming systems.
SplitStream [Castro et al., 2003b] is one of the pioneering research-based
systems for P2P video streaming. As in most early days' systems, it is designed
based on a structured approach in that it uses multicast trees at the applica-
tion level for distributing video packets. Specifically, to build robustness into
the system, a multi-tree approach is used in that the source partitions the
video into multiple stripes and then uses multiple interior-node-disjoint trees
to carry each stripe to the clients. The key property of such a forest of interior-
node-disjoint trees is that each participating node serves as an interior node in
one and only one tree. Consequently, even if a node fails or departs abruptly,
only one stripe of video is affected. Such impairment can be furthered masked
by proper use of redundant coding in the MDC video.
CoolStreaming [CoolStreaming, 2009] represents one of the major pioneer-
ing efforts in large scale video streaming based on P2P technologies [Li et al.,
2007, Sentinelli et al., 2007, Venot and Yan, 2007, Xie et al., 2007, Zhang et al.,
2005b]. The original version was implemented in Python in 2004 and became
an instant hit world-wide. Its design has also inspired many other systems,
including commercial products. Basically, in CoolStreaming, a client starts by
contacting a bootstrap node to obtain a list of currently active peers. It then
randomly selects a subset of such peers as partners. The client can then start
getting required packets from such partners using a standard mechanism, i.e.,
exchanging buffer maps and then getting missing packets from the partners.
One distinctive feature in CoolStreaming is that once a partner, also known