1. E-ODMRP: Enhanced ODMRP with Motion Adaptive Refresh Soon Y. Oh, Joon-Sang Park, Mario Gerla Computer Science Dept. UCLA

2. 2 Multicasting in ad hoc nets  Why multicast in ad hoc nets? Group (1-to-many) communication Wireless “broadcast” medium  ODMRP: On Demand Multicast Routing Protocol One of the most widely used ad hoc multicast routing protocol Simple yet high-performing

3. 3 Join Query Join Reply Forwarding Node Link Multicast Route  On-demand approach: A source initiates JOIN QUERY flooding only when it has data to send  The sender periodically floods JOIN QUERY control messages  All intermediate nodes set up route to sender (backward pointer)  Members send Join Reply messages following backward pointers  Routes from sources to receivers build a mesh of nodes called “forwarding group”. S R R R R Forwarding Group ODMRP: Initialization Phase F F F F

4. 4 Generalize to multiple sources  To make the procedure scalable to large number of sources: stagger “join query” floods aggregate join replies S R R R R S1 S2 Forwarding Group F F F F

5. 5  Source broadcasts data packet to neighbors  Forwarding Group nodes forward multicast packets via “restricted” flooding on the forwarding mesh  Soft state No explicit receiver join/leave messages Forwarding nodes clear state upon timeout  Extremely robust to mobility, fast fading, obstacles, jamming S R R R R S2 Forwarding Group ODMRP: operation

6. 6 Comparison: Packet Delivery Ratio

7. 7 Comparison: O/H = tx/deliver

8. 8 Problem: Forward Group maintenance  Mesh is very resilient to: Short term disruptions (jamming, fading, obstacles) Medium term (connectivity) disruptions, eg FG node moving out of field  FG maintenance To overcome connectivity disruptions, need frequent m esh refresh Short refresh interval (proportional to FG node longevity) needed to keep connectivity in the face of motion  Problem: Short refresh interval leads to high overhead  Refresh rate is a key performance parameter

9. 9  Adaptive route refreshing Route refresh rate is adjusted on-the-fly to environment, i.e., node mobility Adjustment is based on receivers ’ loss reports to source  Local route recovery Receiver estimates packet interval and calculate time out eg. Interval * n If time out expires, the disconnected node proactively grafts onto the FG mesh instead of waiting until next route refresh Solution: motion adaptive refresh + local route recovery

10. 10 Local Route Recovery  Ring search with limited TTL Disconnected node (say node A) floods RECEIVER JOIN locally, e.g. set packet TTL to 1 On reception of RECEIVER JOIN, a Listener node, a neighbor of any forwarder or receiver nodes, sets itself up as a Temporary Forwarders and start forwarding next several data packets Node A sends passive ACKs to one of Temporary Forwarders (say node B) Node B becomes a Forwarder and others clear their status and go back to Listeners

11. 11 Local Route Recovery Source Forwarders ReceiversListeners Receiver Join Data flow A B C D

12. 12 Local Route Recovery A B C D Source Forwarders ReceiversListeners Receiver Join Data flow

13. 13 Local Route Recovery A B C Source Forwarders ReceiversListeners Receiver Join Data flow D

14. 14 Local Route Recovery (Cont.)  If failed Local Recovery, the disconnected node floods entire network with REFRESH REQUEST  On reception of REFRESH REQUEST, sources refresh FG by flooding JOIN QUERY

15. 15 Adaptive Route Refresh  Refresh interval varies between min and max value, e.g. 3 sec and 30 sec  On reception of REFRESH REQUEST (RR), refresh interval is adjusted to: Max > Rfr >Min ( route lifetime/F, 3 sec)  Route lifetime is the time difference between the two events: last JOIN Query arrival and link breakage detection  F is a reduction coefficient, e.g. F=2  If no RR during a refresh interval, linearly and slowly increase refresh interval

16. 16 Passive ACK and Pruning  Intermediate nodes overhear packet transmission from downstream nodes  Data packets serve as passive ACKs  If a Forwarder misses several passive ACKs, it prunes itself from the mesh  Passive ACK suppression technique; a leaf node skips sending a passive ACK if it receives duplicated packets Other node may send a passive ACK A leaf node is changing a upstream forwarder due to mobility

17. 17 Passive ACK Suppression & Pruning Forwarders Receivers Passive ACK Data flow

18. 18 Passive ACK Suppression & Pruning Forwarders Receivers Passive ACK Data flow

19. 19 Passive ACK Suppression & Pruning Forwarders Receivers Passive ACK Data flow

20. 20 Simulation Results  Settings NS2.1b8 100 nodes on 1200x800m 2 Random Way Point mobility model 512 byte/packet Constant bit rate traffic (4 packet/sec) 300 seconds simulation time Scenario 1: Varying mobility  Varying max speed (1 ~ 30m/s) and 0 sec pause time  1 group, 1 source, and 20 receivers

21. 21 Simulation Results (Cont.) Scenario 2: Varying number of receivers  Varying number of receivers (10 ~ 50)  20 m/s max speed and 0 sec pause time Scenarios 3: Varying data rate  Varying data rate 4pkts/sec ~ 30pkts/sec  20 m/s max speed and 0 sec pause time  1 group, 1 source, and 20 receivers Scenarios 4: Varying number of sources  Varying number of sources (1 ~ 6)  20 m/s max speed and 0 sec pause time  1 group and 20 receivers

22. 22 Results in various mobility cases Packet Delivery Ratio E-ODMRP maintains PDR degradation within 1% to ODMRP and surpasses ADMR’s PDR

23. 23 Results in various mobility cases Normalized Packet Overhead E-ODMRP reduces the normalized overhead by 50% to ODMRP’s

24. 24 Results in various group size Packet Delivery Ratio E-ODMRP scales with the number of receivers and shows best PDR with 50 receivers

25. 25 Results in various group size Normalized Packet Overhead E-ODMRP normalized overhead is superior to ODMRP and ADMR

26. 26 Results in various Data Rate Packet Delivery Ratio E-ODMRP outperforms ODMRP and ADMR in high packet sending rate

27. 27 Results in various Data Rate Normalized Packet Overhead E-ODMRP keeps lowest normalized overhead in high packet sending rate

28. 28 Results in various number of Sources Packet Delivery Ratio PDR lines decrease by different factors and E-ODMRP surpasses others when there are more than three sources

29. 29 Results in various number of Sources Normalized Packet Overhead E-ODMRP overhead is near-flat line, but ADMR’s overhead slope suddenly change

30. 30 Conclusion  E-ODMRP : Enhanced ODMRP with motion adaptive refresh  E-ODMRP reduces normalized packet overhead up to 50% yet keeping similar PDR compared to ODMRP  E-ODMRP surpasses ADMR in any case  E-ODMRP achieves high packet delivery ratio with low overhead