05-12-2012, 04:49 PM
A Network Coding Equivalent Content Distribution Scheme for Efficient Peer-to-Peer Interactive VoD Streaming
A Network Coding Equivalent Content.pdf (Size: 1.26 MB / Downloads: 52)
INTRODUCTION
MULTIMEDIA streaming is now a popular Internet
service. However, efficient streaming to a large client
population is hampered by server bandwidth constraints
and the fact that IP-layer multicast is not universally
supported. Peer-to-peer (P2P) collaborative streaming is a
promising solution to the problem of efficiency. In a P2P
system, each peer requests multimedia content from
specific supplying peers. Then, after receiving the data,
the peer caches it in local storage so that the (receiving) peer
can now become a new supplier for other peers. An
important challenge in a P2P collaborative video-ondemand
(VoD) streaming system is to develop an effective
content distribution scheme that can support a dynamic
network among peers, where autonomic peers can join or
leave the system at any time and any place in the network.
The situation is further complicated by the need to support
random access, such as the trick plays of pause/resume,
jump, fast forward (FF), and rewind.
RELATED WORK
Some P2P systems using linear network coding have already
been developed. Wang and Li [7] presented R2, a new
streaming algorithm designed to combine random network
coding with a randomized push algorithm. R2 focuses on
improving the efficiency of live streaming in terms of startup
buffering delays, resilience to the unpredictable behavior of
peers, and bandwidth saving of streaming servers. Wu et al.
[10] proposed some decentralized strategies to eliminate
conflicts among coexisting streaming overlays on contested
bandwidth and combined those strategies with networkcoding-
based content distribution to realize efficient multioverlay
streaming. Chi et al. [11] proposed a buffer-assisted
search (BAS) scheme to increase partner search performance
by decreasing the size of the index structure. They also
designed a novel scheduling algorithm using the dead-line
aware network coding (DNC) to fully utilize network
resources by adopting an appropriate coding window size.
Linear Network Coding of Segments
To reduce the cost of finding parents, the proposed NCECD
scheme, as illustrated in Fig. 3, utilizes linear network coding
technology [6] for media data distribution. When a sufficient
number of encoded blocks are read fromM different parents,
the original segment can be decoded. The proposed NCECD
scheme is a significant advance over the approach shown in
Fig. 2 because it has a much faster parent search time.
Whereas the basic interleaving scheme requires the child to
find a specific parent with the desired block, the NCECD
scheme does not require the search for specific parents.
Application of the NCECD Scheme to P2P VoD
Streaming
In the proposed NCECD scheme, in addition to streaming
the media data of the desired segments, the participant
peers also download other encoded blocks from parent
peers or the server until they have finished downloading N
encoded blocks separately from N segments. At the same
time, participant peers may directly forward these downloaded
blocks to other peers or reencode a received segment
with a random coefficient vector and then forward these
reencoded blocks to maintain high independence among
distributed encoded blocks in the P2P network. An adaptive
incentive mechanism proposed by Habib and Chuang [17]
Simulation Results
1) Startup and jump delays. Startup delay indicates the first
requested segment location latency. For an accurate comparison
of segment location latency, a child peer must always try
to first connect to the closest parent peer in all schemes. For
the competing schemes and the proposed scheme, a child
peer contacts a number of parent peers and begins downloading
media data in parallel. For the competing schemes, a
child peer only needs to find the first parent peer to start
streaming. While the child peer downloads, it continues to
connect to other parent peers for multisource streaming. In
the proposed scheme, a child peer must contact enough
parent peers to start streaming due to the network coding
limitation. However, since all parent peers can be selected,
the proposed NCECD scheme has a higher probability of
connecting to closer parent peers for downloading, and it
does not need to perform a costly search for parent peers
through a segment distribution topology, as used by other
competing schemes.