cs +
−
k
i=1
Φi ≥
−
k
i=1
Θi
. (1)
In other words, the previous inequality simply states that the continuous playout of the stream is possible if the sum of
offered resources (by the ISPs and by the streaming server) is greater than (or equal to) the sum of requested resources.
Furthermore we can say that the previous inequality is the ISP-equivalent of the result presented in [18] (Theorem 1) for
the peers of a P2P streaming system, and the underlying assumptions of Inequality (1) are similar to the ones used in [18]. In
particular, we are assuming that there is a complete connectivity among the ISPs, and that the resource can be split among
the ISPs in a ‘‘continuous manner’’ (in [18] this is called ‘‘fluid assumption’’). In Section 6.1 we discuss the impact of these
simplifying assumptions on realistic P2P streaming systems.
Please note that a peer j ∈ Mi uses several different criteria to select its neighborhood choosing among peers of Mi and
of other ISPs. These impact on the splitting of θj between θ
′
j
and θ
′′
j
. Our goal is the development of a modeling framework
for the study of the interactions among ISPs that fits P2P applications with different characteristics (e.g., trees, multi-tree,
mesh, pull, push, and so on); therefore we do not want to rely on a specific criterion to select peer neighbors (in the same
or other ISPs). We notice that in our framework, the selection criteria change Θi
, i.e., they impact on the rate at which a ISP
requests data to the other ISPs. In other words, we do not account for the selection criteria but we only consider their effects
on the ISP requests. In Section 6.1 we show that our framework does not depend on the particular section method but only
relies on the ISP requests, i.e., Θi
.