1. Trading Coordination For Randomness Szymon Chachulski Mike Jennings, Sachin Katti, and Dina Katabi

2. Wireless mesh networks have high loss rates Roofnet Objective: High throughput despite lossy links Objective: High throughput despite lossy links Avg. 30% loss

3. Use Opportunistic Routing

4. src R1 dst R4 R2 R3 50% 0% 50% 0% 50% Use Opportunistic Routing Best single path  loss prob. 50%

5. In opp. routing [ExOR’05], any router that hears the packet can forward it  loss prob. 0.5 4 = 6% Use Opportunistic Routing Opportunistic routing promises large increase in throughput src R1 dst R4 R2 R3 50% 0% 50% 0% 50%

6. But Overlap in received packets  Routers forward duplicates

7. src R1 dst But R2 Overlap in received packets  Routers forward duplicates P1P1 P2P2 P 10 P1P1 P2P2 P1P1 P2P2

8. src R1 dst But R2 Overlap in received packets  Routers forward duplicates State-of-the-art opp. routing, ExOR imposes a global scheduler: -Requires full coordination; every node must know who received what -Only one node transmits at a time, others listen P1P1 P2P2 P 10 P1P1 P2P2 P1P1 P2P2

9. Global Scheduling? Global coordination is too hard One transmitter src dst

10. Global Scheduling? src dst Global coordination is too hard One transmitter  You lost spatial reuse! Does opportunistic routing have to be so complicated?

11. Our Contributions Opportunistic routing with no global scheduler and no coordination We use random network coding Experiments show that randomness outperforms both current routing and ExOR

12. Go Random Each router forwards random combinations of packets

13. src R1 dst Go Random R2 α P1+ ß P2α P1+ ß P2 γ P1+ δ P2γ P1+ δ P2 Each router forwards random combinations of packets Randomness prevents duplicates No scheduler; No coordination Simple and exploits spatial reuse P1P1 P2P2 P1P1 P2P2

14. src dst1dst2dst3 P1P1 P2P2 P3P3 P4P4 P1P1 P2P2 P2P2 P3P3 P3P3 P4P4 P3 P4 P1 P4 P1 P2 Random Coding Benefits Multicast Without coding  source retransmits all 4 packets

15. src dst1dst2dst3 P1P1 P2P2 P3P3 P4P4 P1P1 P2P2 P2P2 P3P3 P3P3 P4P4 8 P1+5 P2+ P3+3 P48 P1+5 P2+ P3+3 P4 7 P 1 +3 P 2 +6 P 3 + P 4 P3 P4 P1 P4 P1 P2 Random Coding Benefits Multicast Without coding  source retransmits all 4 packets With random coding  2 packets are sufficient Random combinations Random coding is more efficient than global coordination

16. MORE

17. Source sends packets in batches Forwarders keep all heard packets in a buffer Nodes transmit linear combinations of buffered packets srcABdst P1 P2 P3 P1P2P3 =+ b+ ca a,b,c 4,1,3 0,2,1 4,1,3 P1P2P3 =+ 1+ 34 4,1,3P1P2P3 =+ 2+ 10 0,2,1 Can compute linear combinations and sustain high throughput!

18. srcABdst P1 P2 P3 P1P2P3 =+ b+ ca a,b,c 4,1,3 0,2,1 4,1,3 = 2+ 1 0,2,1 8,4,74,1,3 8,4,7 MORE Source sends packets in batches Forwarders keep all heard packets in a buffer Nodes transmit linear combinations of buffered packets

19. Destination decodes once it receives enough combinations o Say batch is 3 packets 1,3,2 5,4,5 4,5,5 P1P2P3 =+ 3+ 21 P1P2P3 =+ 4+ 55 P1P2P3 =+ 5 4 Destination acks batch, and source moves to next batch MORE Source sends packets in batches Forwarders keep all heard packets in a buffer Nodes transmit linear combinations of buffered packets

20. But How Do We Get the Most Throughput? Naïve approach transmits whenever 802.11 allows If A and B have same information, it is more efficient for B to send it Need a Method to Our Madness A B dst

21. Probabilistic Forwarding A B dst

22. Probabilistic Forwarding e1e1 e1e1 e2e2 A B dst Src P1P1 P2P2 Loss rate 50% Loss rate 0%

23. Probabilistic Forwarding e1e1 A B dst Src P1P1 P2P2 e1 ? 50% of buffer e1e1 e2e2 e1 How many packets should I forward?

24. Probabilistic Forwarding e1e1 A B dst Src P1P1 P2P2 50% 0% e1 e1e1 e2e2 Pr = 0.5 Pr = 1 Compute forwarding probabilities without coordination using loss rates

25. A B dst Pr = 0.5 Pr = 1 Can ExOR Use Probabilistic Forwarding To Remove Coordination? P1 P2 P1 Without random coding  need to know the exact packets to forward every time With random coding  need to know only the average amount of overlap Without random coding  need to know the exact packets to forward every time With random coding  need to know only the average amount of overlap Probability of duplicates is 50%

26. MORE needs to adapt to short-term dynamics There are known knowns. These are things we know that we know. There are known unknowns. That is to say, there are things that we know we don't know. But there are also unknown unknowns. Long-term averages are great, but… Wireless is unpredictable over short time-scales

27. Need to balance sent information with received information MORE triggers transmission by receptions A node has a credit counter o Upon reception, increment the counter using forwarding probabilities o Upon transmission, decrement the counter Source stops  No triggers  Flow is done Adapting to Short-term Dynamics

28. Performance

29. Experimental Setup We implemented MORE in Linux 20-node testbed Compare MORE with: o Current Routing (Single Best Path) o ExOR (State-of-the-art Opportunitic Routing) Experiment o Random source-destination pairs o Transmit 5 MB file

30. Testbed 20-node testbed over three floors

31. Testbed 20-node testbed over three floors Avg. loss 27%

32. Does MORE Improve Wireless Throughput? Current MORE 40 80 MORE’s throughput is 2x better than current routing 22% better than ExOR MORE’s throughput is 2x better than current routing 22% better than ExOR Avg. Throughput over 180 src-dst pairs [pkts/s] ExOR

33. Throughput of All Source-Destination Pairs CDF of 180 source-destination pairs Throughput [packets/s] Current MORE ExOR

34. Throughput [packets/s] MORE addresses dead spots 4x4x Zoom in on the worst 10% Current MORE ExOR

35. Sensitivity to Batch Size Throughput [packets/s] ExOR CDF Batch = 8 pkts Batch = 128 pkts MORE CDF Throughput [packets/s] Batch = 8 pkts Batch = 128 pkts MORE works for short flows

36. 2 destinations 3 destinations4 destinations 50 100 What About Multicast? MORE Current +200% +260% +320% MORE improves both multicast and unicast throughput Avg. Throughput Per Destination [pkts/s]

37. MORE for Less! Lesser coordination and lesser rigidity o No scheduler More flexibility o Works on top of 802.11  enjoy spatial reuse o One framework for unicast and multicast More throughput o 22% better than ExOR o 2x better than current routing