1. 1 Increasing the Performance of MANETs Throughput and QoS Performance Enhancing Mechanisms for Unicast and Group Communication in Proactive Mobile Ad Hoc Networks PhD Dissertation Erlend Larsen January 28 th 2011 Erlend Larsen, PhD Dissertation 2011

2. 2 Outline Introduction –Motivation –Challenges –OLSR –Thesis overview Contributions –Unicast routing –Group communication Concluding remarks

3. 3 Motivation Today –Mainly one hop broadcast voice, ”walkie-talkies” or TETRA with low data capacity Tomorrow –Voice –Situational awareness Position sharing Geographically mapped events –Access to maps and construction drawings –Etc. But… Improving the information flow in emergency and rescue operations

4. 4 Challenges Medium access: Contention-based random access –Collisions –interference Distributed routing –Inconsistency, overhead Node mobility Node density –Partitioning or low share of medium access Link quality –Varying Result: Low performance, difficult to support QoS

5. 5 Optimized Link State Routing – OLSR – a proactive routing protocol Maintains a full topology overview Maintains a Connected Dominating Set Many implementations available –Linux, Windows –NS-2 simulator IETF’s proposed proactive routing protocol for MANETs

6. 6 OLSR – Control messages HELLO messages with own neighborhood information periodically broadcasted to all neighbors –Type of link to all neighbors: asymmetrical, symmetrical, lost –MPR selection –Timeout information TC messages –Global link information

7. 7 MultiPoint Relays in OLSR A node selects a subset of its neighbors as MPRs, to reach all 2-hop neighbors MPRs do: –TC generation –Forwarding

8. 8 Overview of the work Thesis structure Unicast routing Rerouting and queueing (Paper A) Gateways and capacity (Paper B) Routing with buffer zones (Paper C) Group communication Preemption mechanisms (Paper D) Optimized SMF (Paper E)

9. 9 Contributions to Unicast Routing

10. 10 Rerouting Time and Queueing in Proactive Ad Hoc Networks Vinh Pham, Erlend Larsen, Knut Øvsthus, Paal Engelstad and Øivind Kure In proceedings of the Performance, Computing, and Communications Conference 2007 (IPCCC 2007), New Orleans, USA, April 11-13, 2007, pp. 160-169.

11. 11 Motivation Discovery: Rerouting due to mobility exceeds the expected 4-6 seconds. SD

12. 12 The contributions Analysis and simulation of the rerouting time Proposed solution of adapting the number of MAC layer retries

13. 13 A link break broken down Link is broken between A and C. A’s queue is being filled up. A transmits data to C New route established via B Last Hello from C received at A Last successfull data transmission from A directly to C Garbage packets are discarded from A’s queue A B C

14. 14 Solution – Adaptive Retry Limit 1 2 3 Packet 1 is transmitted 7 times and discarded Assumes: Retry limit = 7 All packet to the same destination 8 7 9 2 3 7 89 Packet 2 is transmitted 6 times and discarded 7 8 9 Packet 7 is transmitted 1 time and discarded 9 8 In Out Node A’s Interface Queue

15. 15 Results

16. 16 Conclusion Rerouting time is affected by: –Packet size and rate –MAC layer queue size –MAC layer retries Adapting the MAC layer retries reduces the rerouting time.

17. 17 Gateways and Capacity in Ad Hoc Networks Erlend Larsen, Vinh Pham, Paal Engelstad and Øivind Kure In proceedings of the International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services 2008, (I- CENTRIC 2008), Sliema, Malta, October 26-31, 2008, pp. 390-399, ISBN: 978-0-7695-3371-1

18. 18 Motivation Gateways can interconnect ad hoc networks with external networks. The gateway’s position in the ad hoc network may impact the capacity of the ad hoc network Understanding the impact of gateway positions on the offered capacity can be valuable.

19. 19 Investigated scenarios One, two and multiple gateways Traffic flowing either into the network from the gateway to all ad hoc nodes, or vice versa. –Downlink –Uplink With and without dynamic gateway selection

20. 20 One gateway downlink Throughput is highest with the GW near the center. At the center the average number of hops is the lowest. Lack of route is the dominating cause of packet loss

21. 21 Two gateways downlink Throughput greatly increased compared to one gateway. Throughput peak at 750 m separation – where the average number of hops is lowest. The results make a jump at 550 m, i.e. when the two gateways no longer are in each other’s sensing range.

22. 22 MAC layer retransmissions Mobility leads to lower throughput in the downlink scenarios Centered gateway receives network traffic (Uplink scenario)

23. 23 Conclusions 1.The average hop count affects the capacity: –Additional GWs increase the throughput –No further throughput increase when all nodes are in 1-hop range of a GW. 2.The GWs’ sensing range affects the capacity: –Exposed node problem with downlink (from GW to ad hoc nodes) –Hidden node problem with uplink (to GW from ad hoc nodes) 3.Lower throughput for downlink traffic: –Mobility + MAC retransmissions 4.Without dynamic gateway selection, the performance of two gateways equals that of one gateway.

24. 24 Routing with Transmission Buffer Zones in MANETs Erlend Larsen, Lars Landmark, Vinh Pham, Øivind Kure and Paal Engelstad In proceedings of the IEEE International Symposium on a World of Wireless Mobile and Multimedia Networks 2009 (WoWMoM 2009), Kos, Greece, June 15-18, 2009.

25. 25 Motivation SD Paper A: Rerouting due to mobility exceeds the expected 4-6 seconds. Can we anticipate and reroute in advance?

26. 26 Transmission buffer zones S D Buffer zone Safe zone

27. 27 Results

28. 28 Conclusions The buffer zone solution improves the goodput over standard OLSR – even though loops appear more frequently. The size of the buffer zone can be optimized depending on node mobility. The node classification metric may be other than signal strength MAC layer retries Average link loss rate

29. 29 Contributions to Group Communication

30. 30 Preemption Mechanisms for Push-to-Talk in Ad Hoc Networks Erlend Larsen, Lars Landmark, Vinh Pham, Paal E. Engelstad and Øivind Kure Accepted at the 34th IEEE Conference on Local Computer Networks 2009 (LCN 2009), Zürich, Switzerland, October 20-23

31. 31 Motivation Push-to-Talk (PTT) –should be supported in MANETs for emergency and crisis scenarios. –Distributed using multicast/efficient flooding. Without priority, the PTT traffic will be severely impacted by background traffic.

32. 32 The contributions Investigate how PTT traffic is affected by background traffic Study the effect of priority queuing Propose and study three preemption mechanisms –Discard –Buffering –Low priority window Investigate how TCP traffic affects the proposed solutions Save the background traffic

33. 33 Solutions Discard Buffer Low Priority Window Routing layer Interface queue Interface n PbPb PaPa WPbPb n+1n time PaPa …

34. 34 Results Buffer LPW Priority Queuing Mind the gap Discard

35. 35 Conclusions 1.PTT traffic must be protected –priority queuing is not enough. 2.Preemptive discard –effective for PTT –devastating for the background traffic. 3.Buffering and Low Priority Window rescues background traffic –LPW risks reduced priority traffic performance. 4.Preemption initialization is vulnerable –Racing condition with the TCP background traffic

36. 36 Optimized Group Communication for Tactical Military Networks Erlend Larsen, Lars Landmark, Vinh Pham, Øivind Kure, and Paal. E. Engelstad In proceedings of the IEEE Military Communications Conference (MILCOM), San Jose, CA, USA, October 31–November 4, 2010, pp. 1445–1451, ISBN: 978-1- 4244-8179-8.

37. 37 Motivation PTT and SA traffic have different QoS requirements, but must exist simultaneously in the MANET –PTT traffic requires low loss and low latency –SA traffic is more robust Both traffic types may be forwarded using multicast or efficient broadcast SMF using S-MPR showed vulnerability to mobility

38. 38 The contributions Investigating the behavior of S-MPR and NS-MPR –under mobility and varying traffic load. Employing a radio load metric –to select the better algorithm of S-MPR and NS-MPR. Dynamic preemptive choice of forwarding algorithm –to better support the coexistence of PTT and SA data in the network

39. 39 MPR MPR-selector

40. 40 S-MPR

41. 41 NS-MPR

42. 42 SMF forwarding with CF, S-MPR or NS-MPR Classic Flooding – All nodes forward once S-MPR – Source-based MPR forwarding NS-MPR – NON-Source-based MPR forwarding All MPRs forward TopologyCF 30 nodes99.7% 50 nodes98.9% S-MPR 32.1% 19.1% NS-MPR 47.9% 50.7%

43. 43 S-MPR and NS-MPR behavior S-MPR is vulnerable for mobility and collisions NS-MPR results in more transmissions per packet

44. 44 Radio load metric A node can observe the local radio usage and select which algorithm to use for forwarding based on the radio load. –For high loads, S-MPR should be preferred –For low loads, the NS-MPR should be preferred In case of mobility, the radio load will be reduced with S-MPR, making sure more nodes employ NS-MPR. –NS-MPR is better at mobility. The performance of the radio load thresholds lie between those of S-MPR and NS-MPR.

45. 45 Radio load metric

46. 46 Preemptive switch to S-MPR PTT traffic is vulnerable to collisions, and collisions occur before the radio load forces SA traffic to be forwarded using S-MPR A preemptive switch to S-MPR for the SA traffic reduces the impact of SA traffic on the PTT traffic

47. 47 Conclusions Analyzed the behavior of S-MPR and NS-MPR: –NS-MPR – robust, but uses more resources. –S-MPR – vulnerable to mobility. Proposed a radio load metric to switch between S-MPR and NS-MPR distributedly, handling: –High offered load –Mobility Proposed a preemptive forwarding algorithm switch to S- MPR for the SA traffic. –Optimized the performance of the PTT traffic –Allowing the SA service to operate during a PTT session.

48. 48 Concluding Remarks MANETs are exposed to many challenges impacting the performance of the network. Some of the challenges have been addressed through this thesis work: –Node or gateway position –Mobility induced link breaks –Loss due to competing traffic Increased understanding has been provided through this work, and solutions that increase the network performance have been proposed.

49. 49 Thank You!