1. Cristian Lumezanu Dave Levin Neil Spring PeerWise Discovery and Negotiation of Faster Paths

2. Routing Overlays PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 Deployment COST is LOW Potential BENEFIT is HIGH

3. Scalability? PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 Probing should be done more selectively A B C

4. Routing overlays should include an incentive mechanism Fairness? PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 cost benefit > 1 cost benefit < 1

5. PeerWise PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 PeerWise Nodes that can help each other find better paths peer Nodes negotiate and establish pairwise connections to each other Motivation Cost-benefit ratio known before committing any resources Models autonomous system peerings in the Internet Overlays built on self-interest rather than altruism

6. PeerWise PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 60ms A B C D 50ms 30ms 40ms

7. PeerWise properties PeerWise Discovery and Negotiation of Faster Paths HotNets 2007

8. PeerWise properties PeerWise Discovery and Negotiation of Faster Paths HotNets 2007

9. Is mutual advantage common? PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 A B C E G D F Difference between the number of routes each node uses the other for after all peerings have been established PEERING SCORE PeerWise prototype with global knowledge 256 DNS server data set gathered using the King method Each node sends data to all other nodes PeerWise prototype with global knowledge 256 DNS server data set gathered using the King method Each node sends data to all other nodes EXPERIMENT SETUP

10. Is mutual advantage common? PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 100 0 0 74% of the pairs of nodes are happy with existing peerings % of pairs peering score 74 20 

11. PeerWise properties PeerWise Discovery and Negotiation of Faster Paths HotNets 2007

12. Finding shorter detours PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 10ms 100ms Triangle inequality violations indicate the existence of shorter one-hop detours A B C At least 66% of the pairs of nodes in our datasets are long sides in TIVs We use network coordinates to find triangle inequality violations flaws in

13. Network coordinates and TIVs PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 20ms 39ms 62ms A B C 38ms 42ms 26ms B A C InternetMetric space TIVs allowed AC > AB + BC No TIVs AC < AB + BC error(AC) = -36ms Long sides shrink Sum of short sides grows

14. Embedding errors and TIVs PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 The more negative the embedding error of an edge, the higher the probability that the edge is a long side in a TIV The more positive the embedding error of an edge, the higher the probability that the edge is a short side in a TIV …and thus has a one-hop shorter detour …and thus is part of a one-hop shorter detour If long sides shrink If sum of short sides grows

15. PeerWise properties PeerWise Discovery and Negotiation of Faster Paths HotNets 2007

16. Performance PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 % of pairs 100 0 0 300 direct pathbest one-hop path Latency (ms) 100 PeerWise path PeerWise reduces latency by an average of 20%

17. Conclusions PeerWise Discovery and Negotiation of Faster Paths HotNets 2007

18. Future Work PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 Why is there mutual advantage? Extensions for low-loss and failure-free paths Future Work Deployment

19. Are one-hop detours enough? PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 Latency (ms) % of pairs 100 0 0 300 100 75 direct path best one-hop detour path 25% of direct paths are longer than 100ms 3% of detour paths are longer than 100ms 50% improvement in average latency

20. Connectivity PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 % of nodes detour score 100 0 0 60 40 22 Percentage of destinations that a node can reach using its peerings, out of all reachable destinations global knowledge limited knowledge (32 neighbors) 78% of nodes can reach more than 60% of their destinations, with global knowledge 60% of nodes can reach more than 60% of their destinations, with limited knowledge

21. Embedding errors and TIVs PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 The more negative the embedding error of an edge, the higher the probability that the edge is a long side in a TIV The more positive the embedding error of an edge, the higher the probability that the edge is a short side in a TIV …and thus has a one-hop shorter detour …and thus is part of a one-hop shorter detour Percentage of how many times a pair of nodes forms a long side in a TIV out of total number of presences in TIVs TIV SCORE

22. Embedding errors and TIVs PeerWise Discovery and Negotiation of Faster Paths HotNets 2007 0 0 -200 200 embedding error TIV score 0 0 100 50 As the estimation error becomes more negative the nodes form more and more long sides estimated distance – real distance more long sides more short sides