1. Deadline-sensitive Opportunistic Utility-based Routing in Cyclic Mobile Social Networks Mingjun Xiao a, Jie Wu b, He Huang c, Liusheng Huang a, and Wei Yang a a University of Science and Technology of China, China b Temple University, USA c Soochow University, China TEMPLE UNIVERSITY

2. Outline Motivation Problem Solution Simulation Conclusion

3. Motivation Concept : Utility-based routing [Jiewu 12, 13] –Utility is a composite metric Utility (u) = Benefit (b) – Cost (c) –Benefit is a reward for a routing –Cost is the total transmission cost for the routing –Benefit and cost are uniformed as the same unit –Objective is to maximize the (expected) utility of a routing

4. Motivation Concept : Utility-based routing –Valuable message: route (more reliable, costs more) –Regular message: route (less reliable, costs less) senderreceiver message route 1 route 2 route k Benefit is the successful delivery reward

5. Motivation Utility-based routing Cyclic Mobile Social Networks delivery deadline is an important factor for the routing design Deadline-sensitive utility-based routing

6. Problem Cyclic Mobile Social Networks –Example

7. Problem Cyclic Mobile Social Networks –Each cyclic MSN can be seen as a weighted graph –Each edge contains a set of probabilistic contacts –Each probabilistic contact:

8. Problem Deadline-sensitive utility-based routing –Benefit: –Utility: –Expected utility u i (t): the expected utility for node i to send a message to its destination within the deadline t

9. Problem Example: Utility for the successful delivery: u(60)=b-c =20-5=15 Utility for the failed delivery: u(60)=0-c=0-5=-5 Expected utility: u 1 (60)=0.5*15+0.5*(-0.5)=5

10. Problem –Cyclic mobile social network: G=  V, E  V: mobile nodes E: set of probabilistic contacts between nodes T: cycle d: destination –Objective: design a deadline-sensitive utility-based routing algorithm to maximize u i (t) for each node i in V

11. Solution: DOUR Basic idea of DOUR –For single-copy routing –Adopt the opportunistic routing strategy –Nodes iteratively calculate their optimal expected utility values when they encounter –During the iterative computation, each node determines an optimal forwarding sequence –Forward messages according to the optimal forwarding sequence

12. Solution: DOUR Concepts –Forwarding opportunity  , v, p  : the node can send messages to node v at time  with the contact probability p –Forwarding sequence S i (t) An ordered set of forwarding opportunities in the ascending of contact times S i (t) = {   1, v 1, p 1 ,   2, v 2, p 2 , · · ·,   m, v m, p m  } 0 ≤  1 ≤  2 ≤· · ·≤  m ≤ t

13. Solution: DOUR Concepts –Opportunistic forwarding rule each node i forwards messages via the forwarding opportunities in its forwarding sequence in turn, according to the ascending of contact times, until the messages are successfully forwarded to some node, or all forwarding opportunities are exhausted.

14. Solution: DOUR Concepts –Opportunistic forwarding rule Example:

15. Solution: DOUR Concepts –Optimal forwarding sequence the forwarding sequence, through which node i can achieve its optimal expected utility when it forwards messages

16. Solution: DOUR Compute Expected Utility –Theorem 1: Assume that node i has a forwarding sequence S i (t)={   1, v 1, p 1 ,   2, v 2, p 2 , · · ·,   m, v m, p m  }, where the optimal expected utilities of v 1,…,v m are u 1 *(t),…, u m *(t). The expected utility, which is related to this forwarding sequence, satisfies:

17. Solution: DOUR Compute Expected Utility Example:

18. Solution: DOUR Determine Optimal Forwarding Sequence –Determine all forwarding opportunities of node i for the deadline t: O i (t) –For each subset S i (t) of O i (t), we compute the related expected utility according to Theorem 1, until we find the forwarding sequence to maximize this expected utility value

19. Solution: DOUR Determine Optimal Forwarding Sequence –Theorem 2: Let S i (t > τ) denote a subsequence of S i (t), where the contact time of each forwarding opportunity in S i (t > τ) is larger than the time τ. Then, where   j, v j, p j   O i (t).

20. Solution: DOUR Determine Optimal Forwarding Sequence Example:

21. Solution: DOUR Determine Optimal Forwarding Sequence

22. Solution: DOUR The Detailed DOUR Algorithm

23. Solution: DOUR Performance of DOUR –Theorem 3: The iterative computation in DOUR will converge within at most |V| rounds of computation. –Corollary 4: DOUR can achieve the optimal expected utility for each message delivery.

24. Solution: m-DOUR Deadline Sensitive Utility Model for Multi- copy Routing –Each message has multiple copies to be forwarded –If any one copy arrives at the destination before the deadline, the message delivery will achieve a positive benefit as the reward. –If all copies fail to reach the destination, the message delivery will result in zero benefit. –The utility is the benefit minus the forwarding cost of all copies.

25. Solution: m-DOUR Basic Idea of m-DOUR –We only consider the two-hop k-copy routing from the source s to the destination d for a given deadline t –We first derive all forwarding opportunities O s (t) –We let the source s always dynamically select k best forwarding opportunities from O s (t) to transfer messages until all forwarding opportunities are exhausted.

26. Simulation Real Trace Used –UMassDieselNet Trace Algorithms in Comparison –Single-copy routing: DOUR, MaxRatio, MinDelay, MinCost –Multi-copy routing: m-DOUR, Delegation, OOF Metrics Average utility, Delivery ratio, Average delay, Average Cost

27. Simulation Evaluation Settings

28. Simulation Results of Single-copy Routing Algorithms –Average utility vs. Deadline, successful delivery benefit, forwarding cost

29. Simulation Results of Multi-copy Routing Algorithms –Average utility vs. Deadline, successful delivery benefit, forwarding cost

30. Simulation Results –Delivery ratio, Average delay, Average cost

31. Conclusion Our proposed algorithm outperforms the other compared algorithms in utility. Both of the proposed algorithms provide a good balance among the benefit, delay, and cost.

32. Thanks! Q&A