Abstract

Peer-to-peer (P2P) systems are quite attractive due to their ability to deliver large
amounts of data at a reduced deployment cost. They o�er an interesting paradigm
for media streaming applications that can bene�t from the inherent self organization
and resource scalability available in P2P systems.
Recently, some push-pull scheduling strategies have been proposed to replace the
classical pull mechanism in mesh based P2P streaming systems. A push-pull mech-
anism is more e�cient in terms of the overheads and leads to much better playback
delay performance since the pull part is mainly used either at the beginning of the
session or to recover missing content.
In order to exploit such an advantage, we propose to revisit peering strategies
based on the use of a push-pull content retrieval mechanism. Our focus is to minimize
the playback delay experienced by participating nodes. We propose two new peering
strategies that we compare to the state-of-the-art strategies using simulation.
To validate the overlay construction results obtained by simulation, we propose
and solve a linear programming model that proceeds by constructing spanning trees
over the obtained mesh. Such validation leads to the evaluation of the peering strate-
gies independently from the scheduling strategy that is used. The model has a good
scalability and can be extended to re
ect the view of a P2P designer or to �nd the
most ISP-friendly strategy among di�erent overlay construction strategies.
With respect to the content retrieval part, we combine the low scheduling delays
of push scheduling with the resiliency and multi-sender ability of mesh overlays. We
propose a new pure push scheduling strategy, PurePush, where we replace the pull
mechanism by a probabilistic push: Parents of a node push a packet with a relaying
probability to reduce redundancy.
Two variations of PurePush are proposed and compared with respect to playback
delay and redundancy/overhead tra�c to a typical Push-Pull algorithm. PurePush
signi�cantly improves the playback delay experienced by peers in the situation where
there is packet loss.