Introduction - Peer-to-Peer Computing: Applications, Architecture, Protocols, and Challenges - page 2

capability. Indeed, it is now a common practice to have P2P file transfer
through “BlackBerry email service” on mobile phones. As such, wireless P2P
file sharing is not only feasible but also becoming pervasive.
The highly flexible features of P2P computing such as a dynamic popu-
lation (users come and go asynchronously at will, at a dramatic scale, called
flash crowd), dynamic topologies (it is impractical, if not impossible, to en-
force a fixed communication structure), and anonymity, come at a significant
cost-autonomy which, by its very nature, is not always in harmony with tight
cooperation. Consequently, ine cient or lack of cooperation could lead to
undesirable effects in P2P computing. Among them the most critical one is
“free-riding” behavior. Loosely speaking, free-riding occurs when some users
do not follow the presumed altruistic cooperation rules such as sharing files
voluntarily, sharing bandwidth voluntarily, or sharing energy voluntarily, so
as to benefit the whole community.
Such altruistic sharing actions, presumably, would bring indirect and intan-
gible (and even remote) returns to the users. For instance, if everyone shares
files voluntarily, every user would eventually benefit from the high availability
of a large and diverse set of selections. Unfortunately, there are some users
that do not believe or buy in to such utopia-like concepts and would, then,
“rationally” choose to just enjoy the benefits derived from the community, but
not contribute their own resources. Thus, a successful P2P system requires an
effective incentive providing mechanism, which is currently a very hot topic
of P2P research.
Apart from incentives, there are three other major research problems faced
by a P2P system. Firstly, even if a participating peer has all the incentives
to cooperate, there is a trust issue that needs to be handled. Specifically, if
there is no trust management system incorporated in the P2P system, it is
di cult for a cooperative peer to determine whether another remote peer is
trustworthy or not. For example, in a file sharing application, it can be di cult
for a cooperative peer to accept a file sent from a remote peer that may not
be trustworthy.
Secondly, as a P2P system scales up, performance quickly becomes an is-
sue. Indeed, many popular file sharing P2P systems can have hundreds of
thousands of users participating at the same time. The response time per-
ceived by each peer is therefore critically determined by how e cient the
P2P network can deliver the requests and results. One major factor is the
network topology, which governs how the participating peers are connected
among each other. Specifically, P2P networks can have a structured topology,
an unstructured topology, or a hybrid between the two. Nevertheless, for all
P2P systems, topology control is always needed to dynamically adjust the
connectivity among peers in order to optimize the performance of the P2P
applications.
Thirdly, and perhaps most importantly, there is a security issue in practical
use of P2P systems. Indeed, by nature of a P2P system, peers interact without
the intervention of a central authority. Thus, even if incentive and trust are