1. Interdomain Routing as Social Choice Ronny R. Dakdouk, Semih Salihoglu, Hao Wang, Haiyong Xie, Yang Richard Yang Yale University IBC ’ 06

2. Outline Motivation A social choice model for interdomain routing Implications of the model Summary & future work

3. Motivation Importance of Interdomain Routing  Stability  excessive churns can cause router crash  Efficiency  routes influence latency, loss rate, network congestion, etc. Why policy-based routing?  Domain autonomy: Autonomous System (AS)  Traffic engineering objectives: latency, cost, etc.

4. BGP The de facto interdomain routing protocol of the current Internet Support policy-based, path-vector routing  Path propagated from destination  Import & export policy  BGP decision process selects path to use  Local preference value  AS path length  and so on …

5. Previous Studies Policy Disputes (Dispute Wheels) may cause instability [Griffien et al. ‘ 99] Economic/Business considerations may lead to stability [Gao & Rexford ‘ 00] Interdomain Routing for Traffic Engineering [Wang et al. ‘ 05] Design incentive-compatible mechanisms [Feigenbaum et al. ‘ 02]

6. What ’ s Missing Efficiency (e.g., Pareto optimality) Previous studies focus on BGP-like protocols  Increasing concern about extension of BGP or replacement (next-generation protocol)  Need a systematic methodology  Identify desired properties  Feasibility + Implementation Implementation in strategic settings  Autonomous System may execute the protocol strategically so long as the strategic actions do not violate the protocol specification!

7. Our approach - A Black Box View of Interdomain Routing An interdomain routing system defines a mapping (a social choice rule) A protocol implements this mapping Social choice rule + Implementation Interdomain Routing Protocol..... AS 1 Preference AS N Preference AS 1 Route AS N Route

8. Outline Motivation A social choice model for interdomain routing Implications of the model Summary & future work

9. A Social Choice Model for Interdomain Routing What ’ s the set of players?  This is easy, the ASes are the players What ’ s the set common of outcomes?  Difficulty  AS cares about its own egress route, possibly some others ’ routes, but not most others ’ routes  The theory requires a common set of outcomes  Solution  Use routing trees or sink trees as the unifying set of outcomes

10. Routing Trees (Sink Trees) Each AS i = 1, 2, 3 has a route to the destination (AS 0) T(i) = AS i ’ s route to AS 0 Consistency requirement: If T(i) = (i, j) P, then T(j) = P A routing tree

11. Realizable Routing Trees Not all topologically consistent routing trees are realizable  Import/Export policies The common set of outcomes is the set of realizable routing trees

12. Local Routing Policies as Preference Relations How does this work?  Example: The preference of AS i depends on its own egress route only, say, r1 > r2  The equivalent preference: AS i is indifferent to all outcomes in which it has the same egress route  E.g: If T1(i) = r1, T2(i) = r2, T3(i) = r2, then T1 > i T2 = i T3

13. Local Routing Policies as Preference Relations (cont ’ ) Not just a match of theory Can express more general local policies  Policies that depend not only on egress routes of the AS itself, but also incoming traffic patterns  AS 1 prefers its customer 3 to send traffic through it, so T1 > 1 T2

14. Preference Domains All possible combinations of preferences of individual ASes  Traditional preference domains:  Unrestricted domain  Unrestricted domain of strict preferences  Two special domains in interdomain routing  The domain of unrestricted route preference  The domain of strict route preference

15. Preference Domains (cont ’ ) The domain of unrestricted route preference  Requires: If T1(i) = T2(i), then T1 = i T2  Intuition: An AS cares only about egress routes The domain of strict route preference  Requires: If T1(i) = T2(i), then T1 = i T2  Also requires: if T1(i)  T2(i) then T1  i T2  Intuition: An AS further strictly differentiates between different routes

16. Interdomain Social Choice Rule (SCR) An interdomain SCR is a mapping: F: R=(R 1,...,R N )  P  F(R)  A F incorporates the criteria of which routing tree(s) are deemed “ optimal ” – F(R)

17. An example

18. Some Desirable Properties of Interdomain Routing SCR Non-emptiness  All destinations are always reachable Uniqueness  No oscillations possible (Strong) Pareto optimality  Efficient routing decision Non-dictatorship  Retain AS autonomy

19. Protocol as Implementation No central authority for interdomain routing  ASes execute routing protocols Protocol specifies syntax and semantics of messages  May also specify some actions that should be taken for some events  Still leaves room for policy-specific actions <- strategic behavior here! Therefore, a protocol can be modeled as implementation of an interdomain SCR

20. The Model in a Nutshell An interdomain routing system defines a mapping (a social choice rule) A protocol implements this mapping Social choice rule + Implementation Interdomain Routing Protocol..... AS 1 Preference AS N Preference AS 1 Route AS N Route

21. Implications of the Model Some results from literature A case study of BGP from the social choice perspective

22. Some Results from Literature On the unrestricted domain  No non-empty SCR that is non-dictatorial, strategy-proof, and has at least three possible routing trees as outcomes [Gibbard ’ s non-dominance theorem] On the unrestricted route preference domain  No non-constant, single-valued SCR that is Nash-implementable  No strong-Pareto optimal and non-empty SCR that is Nash-implementable

23. A Case Study of BGP BGP..... AS 1 Preference AS N Preference Routing Tree

24. Reverse engineering BGP Non-emptiness: X Uniqueness: X Non-dictatorship: X Unanimity:  Strong Pareto Optimality:  only on strict route preference domain

25. BGP is Manipulable! If AS 1 and 3 follow the default BGP decision process, then AS 2 has a better strategy Following the default BGP decision process is not a Nash equilibrium!

26. Possibility of fixing BGP BGP is (theoretically) Nash implementable (actually, also strong implementable) But, only in a very simple game form The problem: the simple game form may not be followed by the ASes

27. Summary Viewed as a black-box, interdomain routing is an SCR + implementation Strategic implementation requirements impose stringent constraints on SCRs The greedy BGP strategy has its merit, but is manipulable

28. What ’ s next? Design of next-generation protocol (the goal!)  Stability, optimality, incentive-compatibility  Scalability  Scalability may serve as an aide (complexity may limit viable manipulation of the protocol) A specialized theory of social choice & implementation for routing?  What is a reasonable preference domain to consider ?

29. Thank you! Comments or Questions: hao.wang@yale.edu

30. Thank you! Comments or Questions: hao.wang@yale.edu