1. Energy and Spatial Reuse Efficient Network-Wide Real-Time Data Broadcasting in Mobile Ad Hoc Networks B. Tavli and W. B. Heinzelman Julián Urbano jurbano@vt.edu

2. Overview Introduction Background MH-TRACE NB-TRACE Simulation Conclusions

3. Introduction

4. Network-Wide Real-Time Data Broadcasting Military networks –Broadcast –QoS –Can not restrict to single-hop Energy efficiency, efficient spatial reuse and QoS are mandatory –No architecture proposed so far addressing all them –Network-wide Broadcasting through Time Reservation using Adaptive Control for Energy efficiency (NB-TRACE) –Based on MH-TRACE

5. Background

6. Energy Dissipation Different Categories –Transmit mode –Receive mode –Idle mode –Carrier sense mode –Sleep mode

7. Energy Dissipation (II) How to Achieve it? –Unnecessary carrier sensing –Idle energy dissipation –Overhear irrelevant packets –Transmit energy dissipation –Reduce overhead

8. Energy Dissipation (III) Before –IEEE 802.11 supports ATIM Ad Hoc Traffic Indication Message Reduces idle time but doesn’t address overhear Focused on unicast traffic –SMAC Periodically shuts off radios to reduce idle time With low traffic outperforms IEEE 802.11 T SMAC and R SMAC

9. Energy Dissipation (and IV) About overhearing –Information Summarization (IS) packet RTS/CTS packets on top of IEEE 802.11 –Power Aware Multiaccess protocol with Signaling for Ad Hoc Networks (PAMAS) Redundant IS packet? Go sleep! Delay, throughput and transmit dissipation –There is an optimum transmit radio D OP Beyond D OP multi-hop outperforms single-hop Great for constant transmit range radios

10. Efficient Spatial Reuse # retransmissions required for a packet to be received by every node Algorithms –Non-coordinated –Fully coordinated Create a Minimum Connected Dominating Set –Partially coordinated Create a MCDS, almost

11. Efficient Spatial Reuse (II) Non-coordinated –Flooding With Random Access Delay (RAD) from 0 to T RAD –Gossiping With RAD and probability p GSP

12. Efficient Spatial Reuse (III) Fully coordinated algorithms –Based on global info –NP-problem

13. Efficient Spatial Reuse (and IV) Partially coordinated algorithms –Based on local info –Counter-Based Broadcasting (CBB) Count packets until broadcast timer expires If received less than N CBB retransmit –Distance-Based Broadcasting (DBB) Based on received power strength If closest received is beyond D DBB retransmit

14. Quality of Service Necessitates –Low delay # hops traversed and contention level –Low jitter Deviation from periodicity of packet receptions –High Packet Delivery Ratio (PDR) Drops and collisions Parameters –T DROP = 150ms –Packet Generation period (T PG ) –PDR = 95%

15. Quality of Service (and II) Highly related to energy efficiency Centralized Control? –Not practical in Mobile Ad Hoc –High overhead Clustering with Cluster Heads (CH) –Schedule the channel access –Some nodes can sleep

16. MH-TRACE

17. Multi-Hop Time Reservation using Adaptive Control for Energy efficiency

18. MH-TRACE (and II) Gain access through the contention slots If gets access fill out the corresponding IS slot Transmit in the corresponding data slot… …until it finishes? Starvation? Network synchronization through GPS

19. NB-TRACE

20. Design Principles Integrate energy-efficiency in MH-TRACE Flooding –IS = (ID node, ID packet ) –Go sleep! Problems with other algorithms –MH-TRACE is application-based NB-TRACE floods the network and prunes Maintain a Control Dominating Set (CDS)

21. Overview Time Division Multiple Access (TDMA) Initially flood to the whole network ACK the upstream nodes If no ACK in T ACK cease rebroadcast Algorithm –Initial Flooding (IFL) –Pruning (PRN) –Repair Branch (RPB) –Create Branch (CRB) –Activate Branch (ACB)

22. Initial Flooding Broadcast packets to one-hop neighbors Contend channel access and rebroadcast –Eventually every node has received IFL IDD=1 for T IFL so every node wakes up

23. Pruning 3 states for nodes –Passive –Active –Activate Branch (ACB) Problem: stop ACKing from outermost leaf –Eventually, only the source node broadcasts Solution: CHs always rebroadcast –Maintain the Non-Connected Dominating Set

24. Pruning (and II) Eventually 1, 3, 5 and 7 go to passive mode –0, 2, 4 and 6 make up the broadcast tree 5 stops rebroadcast after T ACK, 3 stops after 2T ACK, 1 stops after 3T ACK Problem: the nodes are mobile –Re-flood again? Not efficient

25. Repair Branch Mobility causes CHs to go out and come in –New CH stays in startup mode –Mark the beacon packet –Every node rebroadcasts it Problem: broken trees

26. Create Branch If a node detects an inactive CH in T CRB –Switch to active and rebroadcast

27. Activate Branch If a node does not receive for T ACB –Go to ACB mode –Send ACB packet with p ACB Into the IS slots in order not to modify MH-TRACE –If a node receives an ACB packet Switch to active and begin relying –If there is nothing to send, they go to ACB mode –If an ACB node receives data Switch to active and begin relying

28. Packet Drop Threshold T DROP used throughout the network T DROP-SOURCE used at the source node T DROP-SOURCE =T PG

29. Simulations

30. Overview QoS and energy dissipation on –NB-TRACE –MH-TRACE with Flooding –IEEE 802.11 and SMAC with Flooding Gossiping CBB DBB

31. Environment Data packets of 110bytes Node mobility speed from 0.0 to 5.0m/s –2.5±0.2m/s –2.2 ±0.4m/s 1km wide network 80 nodes Data rate of 32Kbps

32. Performance Analysis 3B = IFL, PRN and RPB 4B = IFL, PRN, RPB and CPB

33. Performance Analysis (II) Time –81.4% in sleep –16.7% in idle –2.8% in transmit, receive and carrier sense 19.4% of the total energy dissipation Energy –82.4% packet transmissions –7.5% IS transmissions –10.1% other control packet transmissions

34. Performance Analysis (III)

35. Performance Analysis (and IV)

36. Varying the Data Rate Adjust the superframe size Adjust # of data slots per frame Superframe time≈T PG =25ms.

37. Varying the Data Rate (and II)

38. Varying the Node Density 1 by 1km network with 48Kbps

39. Conclusions

40. Overview Most of the work to date targeted at deducing transmit energy dissipation only NB-TRACE also targets receiving, idle, sleep and carrier sense dissipation According to the 2 (experimental) energy models, transmit energy is not as dominant as thought

41. Quality of Service Satisfies QoS requirements under several different scenarios –Robustness of the broadcast tree –Maintenance of the NCDS –Cross-layer design –Automatic renewal of channel access

42. Energy Dissipation It is way lower –Coordinated channel access –Packet discrimination –Lower Average Retransmitting Nodes (ARN)

43. Delay It is larger with small networks –Restricted channel access Maintains a regular delay with bigger networks It is much lower with larger networks –High node density –High data rates

44. Jitter Lower to all but MH-TRACE –Channel access granted by CHs after contend

45. Spatial Reuse Better than the others –Robustness of channel access –Full integration with MAC layer –IEEE‘s MAC doesn’t prevent excessive collisions No data!

46. Energy Model Energy savings are related to the model Some radios do not support sleep mode or the dissipation difference is small –However, NB-TRACE performs well

47. Future Work Extend TRACE to multicast and unicast –The blocks are reusable –CHs can become multicasting group members as they always broadcast Realistic environments with channel errors –MH-TRACE is shown to outperform IEEE