Conclusions - Peer-to-Peer Computing: Applications, Architecture, Protocols, and Challenges - page 170

Chapter 8
Conclusions
8.1 Where Are We Now?
P2P applications have been proliferating at an ever increasing rate. To
many people, especially the younger generations, using some kind of P2P ap-
plication is already an essential part of daily life. For instance, people routinely
visit some P2P web sites when they want to download some files (e.g., CD im-
ages, etc.). Furthermore, P2P media applications such as Skype and PPLive
are also the default choices for many people. With the advent of smart-phones,
it is widely envisioned that such P2P applications will have an even higher
penetration as people would very likely use them on the go. Indeed, such de-
velopment trends are major motivating factors for even large corporations to
consider a P2P contents delivery model.
By and large, P2P architectures can still be generally classified into struc-
tured and unstructured types. As we have discussed in Chapter 3, both types
have their merits and drawbacks. Yet the current trend is that unstructured
architectures play a more dominant role because of its inherent robustness
and higher scalability. Structured architecture, on the other hand, is mainly
used as an auxiliary component for meta-data indexing, e.g., locating a par-
ticular tracker server. This trend will likely continue as P2P applications are
increasingly used in a wireless and mobile manner.
As discussed in Chapter 4, in order to maintain a specific data sharing
architecture, some pre-defined topology control actions have to be taken by
participating peers. This is all the more important for structured architec-
tures, which, indeed, have stringent requirements for participants. However,
in view of the possibly vigorous peer dynamics, demanding topology control
actions might be a nuisance for the peers. This is one of the reasons why an
unstructured architecture is more appealing in a large scale P2P system with
dynamically changing peer population (i.e., churn). This is fundamentally re-
lated to the autonomous and rationally selfish nature of a peer.
A P2P system, at a fundamental level, is nothing more than a dynamic
“organism” constituted by a large number of self-optimizing peers. As such,
a P2P system's “survival” (in a holistic sense) largely depends on the cooper-
ation among peers. Yet, each peer, being rationally selfish, contributes to the
community by “accident” rather than “on purpose.” Thus, to ensure that the
171