1. An Overview of Peer-to-Peer Networking CPSC 441 (with thanks to Sami Rollins, UCSB)

2. Outline P2P Overview –What is a peer? –Example applications –Benefits of P2P P2P Content Sharing –Challenges –Group management/data placement approaches –Measurement studies

3. What is Peer-to-Peer (P2P)? Napster? Gnutella? Kazaa? Most people think of P2P as music sharing (but it can also be used for good purposes! :)

4. What is a peer? Contrast to Client-Server model –Servers are typically well-resourced, and centrally maintained and administered –Client has fewer resources than a server P2P: nodes are “equals”

5. What is a peer? (cont’d) A peer’s resources are similar to those of the other participants P2P peers communicate directly with each other and share resources Typically at App Layer (ignorant of physical network topology)

6. P2P Goals/Benefits Cost sharing Resource aggregation Improved scalability/reliability Increased autonomy Anonymity/privacy Ad-hoc communication

7. P2P Application Taxonomy P2P Systems Distributed Computing SETI@home File Sharing Gnutella Collaboration Jabber Platforms JXTA

8. P2P File Sharing Approaches Centralized Flooding Document Routing

9. Centralized Napster model Benefits: –Efficient search –Limited bandwidth usage –No per-node state Drawbacks: –Central point of failure –Limited scale –Copyright/legal issues BobAlice JaneJudy

10. Flooding Gnutella model Benefits: –No central point of failure –Limited per-node state Drawbacks: –Slow searches –Bandwidth intensive Bob Alice Jane Judy Carl

11. Document Routing FreeNet, Chord, CAN, Tapestry, Pastry model Benefits: –More efficient searching –Limited per-node state Drawbacks: –Limited fault-tolerance vs redundancy 001 012 212 305 332 212 ?

12. Current Research Peer discovery, group management, data location and placement –Chord, CAN, Tapestry, Pastry Security, privacy, anonymity, trust –Publius, FreeNet Reliable, efficient file exchange Performance studies –Gnutella measurement study

13. Management/Placement Challenges Per-node state Bandwidth usage Search time Fault tolerance/resiliency

14. Document Routing – Chord MIT project Uni-dimensional ID space Keep track of log N nodes Search through log N nodes to find desired key N32 N10 N5 N20 N110 N99 N80 N60 K19

15. Cost Comparisons log b NNeighbor map Pastry b log b N log b NGlobal MeshTapestry 2ddN 1/d Multi- dimensional CAN log N Uni- dimensional Chord StateSearchModel b log b N + b

16. Remaining Problems? Hard to handle highly dynamic environments Usable services Methods don’t consider peer characteristics

17. Measurement Studies “Free riding” on Gnutella Most studies focus on Gnutella Want to determine how users behave Low success rates for transfers (30%?) Recommendations for the best way to design systems

18. Free Riding Results Who is sharing what? August 2000 The topShareAs percent of whole 333 hosts (1%)1,142,64537% 1,667 hosts (5%)2,182,08770% 3,334 hosts (10%)2,692,08287% 5,000 hosts (15%)2,928,90594% 6,667 hosts (20%)3,037,23298% 8,333 hosts (25%)3,082,57299%

19. Saroiu et al Study May 2001 Napster crawl –query index server and keep track of results –query about returned peers –don’t capture users sharing unpopular content Gnutella crawl –send out ping messages with large TTL

20. Results Overview Lots of heterogeneity between peers –Systems should consider peer capabilities Peers don’t always tell the truth! –Systems must be able to verify reported peer capabilities or measure true capabilities

21. Reported Bandwidth

22. Measured Bandwidth

23. Measured Latency

24. Connectivity

25. Conclusion P2P is an interesting and useful model Soon will be the dominant part of Internet traffic volume (if it isn’t already!!) There are lots of technical challenges to be solved (scalability, security, caching, …)