Michele Garetto - unito Daniel R. Figueiredo - EPFL Rossano Gaeta - unito Matteo Sereno - unito

1. IFIP – Performance 2007 A Modeling Framework to Understand the Tussle between ISPs and Peer-to-Peer File Sharing Users Michele Garetto - unito Daniel R. Figueiredo - EPFL Rossano Gaeta - unito Matteo Sereno - unito

2. P2P is starting to dominate Internet Traffic P2P represented 60% of Internet Traffic at the end of 2004

3. Impact on Service Providers P2P is driving consumer broadband uptake …and broadband is driving P2P uptake • P2P is the dominant protocol • In excess of 92% of P2P traffic crosses transit/peering links • P2P protocols will aggressively consume any available bandwidth capacity • Due to P2P’s symmetrical nature on average 80% of upstream capacity is consumed by P2P • P2P affects QoS levels for ALL subscribers • Service Providers can not afford to block or restrict P2P • ISPs must intelligently manage P2P - blocking and shaping doesn’t work

4. The ISP perspective vs P2P:threat or opportunity ? • P2P traffic: friend or foe ? • friend: driving force for adoption of broadband access by the users • foe: overwhelming amount of traffic • What is the best strategy to manage P2P traffic in my network ? • Try to kill it ? • Do nothing ? • Educate it ? How ?

5. Strategies to manage P2P traffic • Acquire more bandwidth • Block P2P traffic • Traffic shaping (e.g., priority to non-P2P) • Pricing schemes based on user traffic volumes / bandwidth caps • Network caching / customized P2P application within ISP • Application-layer redirection of P2P traffic

6. Our contribution • Simple model to analyze the impact of P2P traffic on an ISP network • Basic insights about system dynamics • Performance evaluation tool to evaluate different strategies to manage P2P traffic and guide ISP choices

7. Network scenario User issues a query Solved outside  consumes bandwidth Bd Solved inside  does not use bandwidth Bd • Exploitation of “traffic locality”

8. System model object retrieval probability:

9. Minimum service level for user i User Utility Function • We express the utility of user i as: probability of successful object retrieval shape parameter subscription cost User utility 1 0.5 σ 0 1 0 0.2 0.4 0.6 0.8 -0.5 -1

10. ISP Utility Function • We express the utility of the ISP as: cost per unit of external bandwidth fixed charge profit from subscribers’ fee • The ISP starts the service only if

11. Modeling traffic locality • Probability that there exist at least one internalcopy of an object replicated r times in the system Number of internal copies Number of external copies • Probability to download from internal (existing) copy (i.e., exploit traffic locality) :

12. Modeling object replication • We need to compute the number r of replicas of an object at the time the object is requested, taking into account: • Different popularity of contents • Temporal evolution of object popularity • Introduction of new contents • Cancellation of replicas by the users • We have developed an analytical technique based on Poisson shot noise processes

13. Modeling object replication • Example: (2 objects) popularity A video from the news A popular song time t1 t2 request

14. Results • Assumption: n identical users • N = 50 millions • request rate by user (object/day) • introduction of new contentsby user (object/day) Minimum required external bandwidth:

15. The impact of object replication (r) 5000 Minimum required bandwidth Bmin (objects/day) r = 500 r = 1000 4000 r = 1500 3000 2000 1000 0 0 10000 20000 30000 40000 Number of users, n

16. The impact of efficacy in exploiting traffic locality () 100000 Minimum required bandwidth Bmin (objects/day) = 0.25 = 0.5 80000 = 0.75 = 1.0 60000 40000 20000 0 0 100000 300000 500000 700000 Number of users, n

17. nmin Impact of subscription cost, c 40000 c = 0.25 UISP Utility of ISP c = 1.0 30000 c = 1.4 20000 c = 1.6 10000 0 -10000 0 5000 10000 15000 20000 25000 30000 Number of users, n

18. Critical mass of users, nmin 140000 β2 = 6 nmin β2 = 5 120000 Cost per unit of bandwidth β2 = 4 100000 β2 = 3 80000 60000 40000 20000 0 0 200 400 600 800 1000 Average object replication, r

19. Model refinements • Impact of finite bandwidth of the users

20. Model refinements • Impact of finite bandwidth of the users • In case of constant traffic load, the cost for the ISP increases if it provides more download bandwidth to the users (!) • The system performance is strongly affected by the upload bandwidth of the users, thatshould be larger than or equal to the download bandwidth (contrary to common practices, e.g., ADSL lines !)

21. Impact of asymmetric access bandwidths (for fixed number of users = 20000) 12000 Minimum required bandwidth Bmin (objects/day) bd = 0.1 bu bd = bu 10000 bd = 2 bu bd = 3 bu 8000 bd = 4 bu 6000 4000 2000 0 0 500 1000 1500 2000 2500 3000 3500 4000 Upload bandwidth, bu

22. The impact of bd 7000 bd = 2000 Minimum required bandwidth Bmin (objects/day) bd = 1000 6000 bd = 500 bd = 100 5000 bd = 10 4000 3000 2000 1000 0 0 5000 10000 15000 20000 25000 30000 Number of users, n

23. Conclusions • We have developed a simple analytical model to: • analyze the interaction between P2P traffic and ISP • understand the impact of several parameters • guide the selection of the optimal strategy to manage P2P traffic • Future work • More model refinements • Multiple ISPs competing with each other

24. Questions ? Comments ?