doi:10.1016/j.peva.2007.06.003
Copyright © 2007 Elsevier Ltd All rights reserved.
A modeling framework to understand the tussle between ISPs and peer-to-peer file-sharing users
aDipartimento di Informatica, Università di Torino, Torino, Italy
bPESC/COPPE Federal University of Rio de Janeiro, Rio de Janeiro, Brazil
Available online 20 June 2007.
Abstract
Recent measurement studies have shown that traffic generated by peer-to-peer (P2P) file-sharing applications has started to dominate the bandwidth consumption on Internet access links. The prevailing use of P2P applications carries with it significant implications for Internet Service Providers (ISPs): on the one hand increased levels of P2P traffic result in additional costs for an ISP, which has to provide a satisfactory service level to its subscribers. On the other hand, P2P applications are a major driving force for the adoption of broadband access, which is a significant source of revenue for the ISPs. A successful strategy to manage P2P traffic must address both the ISP perspective of costs and the subscriber perspective of quality of service. While several practical solutions have been identified to manage P2P traffic in a network, no analytical studies have been proposed so far to evaluate their effectiveness in specific contexts. In this paper we propose a modeling framework that allows the optimal strategy to be identified for an ISP as a function of the several factors that come into play. In particular, our model shows that P2P-friendly solutions become lucrative when the ISP can attract a sufficiently large number of subscribers. Our modeling framework also illustrates several other interesting phenomena that occur in the tussle between the ISP and its subscribers.
Keywords: Peer-to-peer sharing systems; Analytical modeling; Traffic management; Utility-based modeling
Fig. 1. Network scenario.
Fig. 2. Simple model for the outcome of queries generated by ISP subscribers.
Fig. 3. The utility function Ui for α=5 and c=1.
Fig. 4. Bmin as a function n, for different values of r (upper left plot); Bmin as a function of n, for different values of γ (upper right plot); Bmin as a function of n, for different values of q (lower left plot); Bmin as a function of n, for different values of 
(lower right plot).
Fig. 5. 
as a function of n, for different values of c (upper-left plot); optimum c as function of n, and corresponding maximum 
(upper-right plot); 
as a function of n, for different values of q (lower-left plot); nmin as a function of r, for different values of β2 (lower-right plot).
Fig. 6. Example of evolution of popularity for two different objects.
Fig. 7. Probability pg that a user keeps doing a data transfer as a function of the initial achieved rate b.
Fig. 8. Refined model for the outcome of queries generated by ISP subscribers.
Fig. 9. Peer selection probability as a function of number of replicas, for ν=0.8.
Fig. 10. Peer selection probability as a function of number of replicas, for ν=1.2.
Fig. 11. Bmin as a function n, for different values of η, bd=1000 (left plot); Bmin as a function of n, for different values of bd, η=4 (right plot).
Fig. 12. Ccdf of the number of replicas of a requested object in the whole P2P community.
Fig. 13. Conditional pdf of the number of online replicas within the ISP network.
Fig. 14. Bmin as a function bu=bd, for different values of the impatience parameter η.
Fig. 15. Bmin as a function of bu for η=0.5 and different values of bd.
Fig. 16. Bmin as a function of bu=bd, for different values of γ.
Fig. 17. Maximum σ as a function of bu=bd, for different values of ν.
Table 1.
Native parameters
This work has been partially supported by the Italian Ministry for University and Research (MIUR) within the frameworks of the FAMOUS (PRIN) and of the PROFILES (PRIN) projects.
Corresponding author. Tel.: +39 011 6706718; fax: +39 011 751603.
Abstract
Purchase PDF (1483 K)
E-mail Article
Add to my Quick Links
Cited By in Scopus (1)
Purchase PDF (371 K)
