Topology Control - Peer-to-Peer Computing: Applications, Architecture, Protocols, and Challenges - page 56

C1 C2 C1 C2

C3 C4 C3 C4

inefficient
overlay topology

efficient
overlay topology

C1 C4

C2

physical
network
topology

C3

FIGURE 4.1: An illustration of the topology mismatch problem: unneces-
sary and ine cient overlay connections are made between peers [Liu et al.,
2005b].

connecting to a peer with more resources in terms of file data owned, upload
bandwidth, etc. Connectivity is also an important consideration because con-
necting to a high-degree peer presumably allows for better reachability in the
network.
However, as mentioned above, if a topology control scheme is designed from
a global perspective only, it may not work very well in practice because peers
are inherently selfish in the sense that it might not be its interest in optimizing
some global performance measures, despite that such “altruistic” actions in
fact will benefit the peer in the long run also. Thus, it is important to attack
the topology control problem in a fully distributed manner using a game
theoretic perspective. Specifically, it might be more practical to formulate
some game strategies for each peer so that local (or selfish) optimizations also
lead to global performance enhancements.
As elaborated in this chapter below, topology control in a P2P system is
still a largely open research topic and has only recently got intensive attention.
We first describe a highly general framework that is adopted by many
recent topology control schemes. Specifically, in this framework, the topol-
ogy control actions are fully distributed—relying only on peers' local actions.
As a result, the topology control schemes are highly practical. However, the
convergence of the topology control actions might take a long time.
We then survey some recently designed techniques for topology control
in a structured P2P network. This is followed by a detailed survey of many
interesting unstructured topology control schemes.