Topology Control - Peer-to-Peer Computing: Applications, Architecture, Protocols, and Challenges - page 71

bile users. The term community is defined as a set of two or more mobile users
that perform in a particular habit. For example, in a music file sharing net-
work, users with similar preference and fans of similar idols are recognized
as members of the same “community.” Members from the same community
follow the same wakeup schedules. The rationale behind this is that file shar-
ing is usually carried out between users with similar interest (this is the usual
reason why a sender owns the favorite file that the requester is asking for). Us-
ing community formation the topology control system not only increases the
chance of getting a file but also allows a group of nodes to sleep and conserve
energy when other communities are active.
4.7 Case Study: PPLive
As discussed in Chapter 3, PPLive exhibits a random graph structure [Hei
et al., 2007a, Vu et al., 2010]. From a topology control perspective, peers in
the PPLive system are pretty autonomous. Connectivity is mainly handled by
the usual approach—peers obtain neighbor lists from their respective channel
management servers (a.k.a. trackers), and then attempt connecting with peers
on the received lists. According to the measurement study conducted by Vu
et al. [Vu et al., 2010], peers in PPLive incline to choose neighbors that are
topologically closer. In other words, the peer selection process seems to be
locality aware. Yet perhaps this is due to the fact that a predominantly large
population of PPLive users are in China. One the other hand, for the temporal
dimension, PPLive peers are found to be quite impatient [Vu et al., 2010].
Specifically, in Vu et al.'s performance study [Vu et al., 2010], about 50% of
sessions are shorter than 10 minutes. Consider that y is the probability that
a node's session length is 10×x minutes. Vu et al. devised a mathematical
model of the session time as: y = ae 10bx , where a and b are some constants
(with a > 0 and b < 0). The degree of peer churn is thus quite high.
4.8 Summary
Topology control is critical for a practical P2P system to deliver good
performance in a resilient manner. Specifically, peers must carry out proper
and timely topology change actions in response to variations in network con-
nectivity situations due to peers joining or departure. However, such actions
have to be localized; otherwise, a great overhead is needed that may actually
aggravate the network connectivity changes.