1. 1 Audio Streaming over Bluetooth Scatternet: using Adaptive Link Layer Team members: Sewook Jung, Jungsoo Lim, Soon Young Oh Tutor: Ling-Jyh Chen Professor Mario Gerla CS218 – Fall 2003

2. 2 Outlines Background Background Adaptive Automatic Retransmission ReQuest (ARQ) Retransmission Timeout (RTO) Adaptive Automatic Retransmission ReQuest (ARQ) Retransmission Timeout (RTO) Previous Research Previous Research Related work Related work Implementations Implementations Simulations Simulations Conclusion Conclusion Future Work Future Work

3. 3 Background Multimedia contents are prosperous Multimedia contents are prosperous –Eg. MP3 audio Wireless Personal Area Network (PAN) needs to support multimedia Wireless Personal Area Network (PAN) needs to support multimedia The varying nature of the wireless link can make streaming over wireless a challenging problem The varying nature of the wireless link can make streaming over wireless a challenging problem Packets are arrived to client with a consistent rate Packets are arrived to client with a consistent rate

4. 4 Background (cont’d) ARQ mechanism ARQ mechanism –Packets being dropped/delayed in bad link –Beneficial to non-real-time traffic –Need modifications for real-time/streaming traffic ARQ retransmission limit ARQ retransmission limit –Too high Packets are severely delayed Packets are severely delayed Streaming audio/video quality is degraded Streaming audio/video quality is degraded –Too low large number of packets are dropped at the link layer large number of packets are dropped at the link layer Also causes poor audio quality. Also causes poor audio quality.

5. 5 An Adaptive ARQ RTO Original Bluetooth Original Bluetooth –stop-and-wait ARQ scheme at link layer packet is retransmitted until receives ACK or retransmission timeout (RTO) is exceeded. packet is retransmitted until receives ACK or retransmission timeout (RTO) is exceeded. –In most current Bluetooth chipsets the default RTO is infinite the default RTO is infinite To provide reliable link. To provide reliable link. –Infinite RTO degrades real-time streaming audio/video quality

6. 6 Previous Research Fixed ARQ RTO Fixed ARQ RTO –Use a fixed finite RTO –Impossible to accommodate all different link qualities with one fixed value. Adaptive ARQ RTO Adaptive ARQ RTO –Adjust RTO by measurement of previous RTT –Improvement on average delay time and the packet success rate RTT increase --  decrease ARQ RTO RTT decrease --  increase ARQ RTO

7. 7 Previous Research (cont’d) The RTO equation SRTT’ = (1-  ) X SRTT +  X RTT(1)  X RTO; if RTT < SRTT(2) RTO’ =  X RTO; if RTT > SRTT RTO; if previous packet is dropped SRTT = smooth RTT,  = 1.1 β = 0.9  = 0.25

8. 8 Previous Research (cont’d) Set the upper bound and lower bound for ARQ RTO Set the upper bound and lower bound for ARQ RTO RTO min = 2 X T packets (= 6*625ms in DH5) RTO min = 2 X T packets (= 6*625ms in DH5) RTO max = T packets X Max(Available Buffer X 75%, 2) RTO max = T packets X Max(Available Buffer X 75%, 2) T packet = time interval between first packet fragments and last fragments’ ACK Available buffer = (system maximum input buffer – used buffer)/packet size

9. 9 Previous Research (cont’d) Adaptive ARQ RTO Results Adaptive ARQ RTO Results –Enhance the streaming audio quality remarkably –Robust solution for real-time/streaming data over wireless network.

10. 10 Related work TCP-Friendly Rate Control (TFRC): equation based TCP rate control TCP-Friendly Rate Control (TFRC): equation based TCP rate control Video Transport Protocol (VTP): sender adjust the sending rate based on estimated eligible rate Video Transport Protocol (VTP): sender adjust the sending rate based on estimated eligible rate RAP: End-to-end Rate Based Control: mimics TCP’s AIMD behavior RAP: End-to-end Rate Based Control: mimics TCP’s AIMD behavior RCS: A Rate Control Scheme: source probes the connection with dummy packets, and adjust sending rate RCS: A Rate Control Scheme: source probes the connection with dummy packets, and adjust sending rate

11. 11 Implementations Blueware: Blueware: –Developed by MIT –Bluetooth simulator as an extension to NS –Various Scatternet formation and link scheduling schemes.

12. 12 Implementations (cont’d) Applications L2CAP Bluetooth Radio Host Controller Interface Bluetooth Baseband LMP Bluetooth Stack

13. 13 Implementations (cont’d) Topology formation Manipulate the topology formation Manipulate the topology formation –Set position of nodes manually –Original Blueware has only random topology formation The examples of topology formations: 1 hop 2 hops3 hops 2 flows3 flows

14. 14 Implementations (cont’d) Original Method Application L2CAPL2CAP HCI/LC Receiver Layer QueueLayer Layer RTT < RTO RTT Partial RTT

15. 15 Implementations (cont’d) Original Method Application L2CAPL2CAP HCI/LC Receiver Layer QueueLayer Layer Partial RTT > RTO RTT Partial RTTHCI_FLUSH

16. 16 Implementations (cont’d) Next Packet Drop Application L2CAPHCI/LCReceiver Layer LayerLayer RTT1 < RTO RTT1 RTT2

17. 17 Implementations (cont’d) Next Packet Drop Application L2CAPHCI/LCReceiver Layer LayerLayer RTT1 > RTO RTT2 < RTO RTT1 RTT2

18. 18 Implementations (cont’d) Flow Control Application L2CAPL2CAP HCI/LC Receiver Layer QueueLayer Layer RTT < RTO RTT

19. 19 Implementations (cont’d) Flow Control Application L2CAPL2CAP HCI/LC Receiver Layer QueueLayer Layer RTT > RTO RTT

20. 20 Implementations (cont’d) Flow Control Application L2CAPL2CAP HCI/LC Receiver Layer QueueLayer Layer RTT < RTO Drop Queue size = 5 RTT

21. 21 Implementations (cont’d) Generating Packet Error Generating Packet Error –Blueware supports packet error rate (PER) instead of bit error rate (BER) –DH5 mode is used for all RTP packets where packet size is 2712 bits and a packet length is five Bluetooth slots –PER is defined as P = 1 – (1 – b) s b = bit error rate, s = packet size

22. 22 Implementations (cont’d) Generating Packet Error (Cont’d) Generating Packet Error (Cont’d) Burst Errors Burst Errors –once the error starts, the probability of having an error in the next bit is extraordinarily high such as 90%. –If the burst error occurs in the middle of the packet, it may not affect the next packet. –However, if it occurs at the end of the packet, there is a great probability of affecting the next packet.

23. 23 Implementations (cont’d) Goo d Bad Pbg Pgb Pbb Pgg Burst Error transition diagram Bit error rate: Pgg: 1-BER Pgb: BER Pbb: 0.9 Pbg: 0.1

24. 24 Experiment Results Adaptive RTO

25. 25 Experiment Results (cont’d) Throughput of Next packet drop (2nodes) Throughput of Next packet drop (2nodes)

26. 26 Experiment Results (cont’d) Delay of Next packet drop (2nodes) Delay of Next packet drop (2nodes)

27. 27 Experiment Results (cont’d) Throughput of Flow control (2nodes) Throughput of Flow control (2nodes)

28. 28 Experiment Results (cont’d) Delay of Flow control (2nodes) Delay of Flow control (2nodes)

29. 29 Experiment Results (cont’d) Packet Success Rate with 2 Nodes Packet Success Rate with 2 Nodes

30. 30 Experiment Results (cont’d) Packet Success Rate with 3 Nodes Packet Success Rate with 3 Nodes

31. 31 Experiment Results (cont’d) Packet Success Rate with 5 Nodes Packet Success Rate with 5 Nodes

32. 32 Experiment Results (cont’d) Fairness Fairness –Topology –Fairness in 2 flows topology –Unfairness in 3 flows topology 2 flows3 flows

33. 33 Experiment Results (cont’d) 2 Flows (Adaptive RTO : Next packet drop)

34. 34 Experiment Results (cont’d) 3 Flows (Adaptive RTO : Next packet drop)

35. 35 Experiment Results (cont’d) 3 Flows (No RTO)

36. 36 Experiment Results (cont’d) Success Rate of Random Error vs. Burst Error

37. 37 Conclusion Success rates were about the same among next packet drop, flow control, and fixed RTO approach Success rates were about the same among next packet drop, flow control, and fixed RTO approach Next packet drop method improved average delay, but throughput suffered Next packet drop method improved average delay, but throughput suffered Flow control method did not improve throughput nor delay Flow control method did not improve throughput nor delay Unfairness detected in 3 flow topology Unfairness detected in 3 flow topology Negligible difference in experiment results between the bit error model and the burst error model Negligible difference in experiment results between the bit error model and the burst error model

38. 38 Future Work Intelligent HCI_FLUSH Intelligent HCI_FLUSH –Previous HCI_FLUSH deletes packets based on connection_handle –All packets contain connection_handle information connection_handle information –HCI packet header or baseband header cid information cid information –L2CAP header –Remove packets which have specific connection_handle or cid Intelligent RTO Intelligent RTO –Adjust RTO based on jitter –New RTO equation: jitter = RTT – T packets ( = 6 *625ms in DH5) jitter = RTT – T packets ( = 6 *625ms in DH5) RTO = RTO - jitter RTO = RTO - jitter –RTT > T packets RTO decrease –RTT < T packets RTO increase

39. 39 Future Work (cont’d) Combination of Adaptive Packet Type (APT) and Adaptive RTO Combination of Adaptive Packet Type (APT) and Adaptive RTO –Combine adaptive RTO scheme with adaptive packet type (i.e. DH5, DH3, DH1, DM5, DM3, DM1) –Choose the best packet type for different BER ranges –Implement the functionality to the Bluetooth LC layer –Optimal packet type can be selected dynamically

40. 40 References J.C. Haartsen, " The Bluetooth Radio System," IEEE Personal Communications Magazine, Feb. 2000. J.C. Haartsen, " The Bluetooth Radio System," IEEE Personal Communications Magazine, Feb. 2000.The Bluetooth Radio SystemThe Bluetooth Radio System NS2 Simulator: http://www.isi.edu/nsnam/ns/ NS2 Simulator: http://www.isi.edu/nsnam/ns/http://www.isi.edu/nsnam/ns/ L.-J. Chen, R. Kapoor, K. Lee, M. Y. Sanadidi, M. Gerla, " Audio Streaming over Bluetooth: An Adaptive ARQ Timeout Approach," L.-J. Chen, R. Kapoor, K. Lee, M. Y. Sanadidi, M. Gerla, " Audio Streaming over Bluetooth: An Adaptive ARQ Timeout Approach,"Audio Streaming over Bluetooth: An Adaptive ARQ Timeout ApproachAudio Streaming over Bluetooth: An Adaptive ARQ Timeout Approach Reza Rejaie, Mark Handley, Deborah Estrin, " RAP: An End-to-end Rate-based Congestion Control Mechanism for Realtime Streams in the Internet," In Proceedings of IEEE INFOCOM 1999. Reza Rejaie, Mark Handley, Deborah Estrin, " RAP: An End-to-end Rate-based Congestion Control Mechanism for Realtime Streams in the Internet," In Proceedings of IEEE INFOCOM 1999. RAP: An End-to-end Rate-based Congestion Control Mechanism for Realtime Streams in the Internet RAP: An End-to-end Rate-based Congestion Control Mechanism for Realtime Streams in the Internet G. Holland, and N. Vaidya," Analysis of TCP performance over mobile ad hoc networks," In Proceedings of ACM Mobicom'99, Seattle, Washington, 1999. G. Holland, and N. Vaidya," Analysis of TCP performance over mobile ad hoc networks," In Proceedings of ACM Mobicom'99, Seattle, Washington, 1999. Analysis of TCP performance over mobile ad hoc networks Analysis of TCP performance over mobile ad hoc networks Balk, D. Maggiorini, M. Gerla, and M. Y. Sanadidi, " Adaptive MPEG-4 Video Streaming with Bandwidth Estimation, ", UCLA. Balk, D. Maggiorini, M. Gerla, and M. Y. Sanadidi, " Adaptive MPEG-4 Video Streaming with Bandwidth Estimation, ", UCLA.Adaptive MPEG-4 Video Streaming with Bandwidth EstimationAdaptive MPEG-4 Video Streaming with Bandwidth Estimation J. Tang, G. Morabito, I. F. Akyildiz, and M. Johnson, "RCS: A Rate Control Scheme for Real-Time Traffic in Networks with High Bandwidth-Delay Products and High Bit Error Rates," In Proceedings of Infocom 2001, Anchorage, AK, 2001. J. Tang, G. Morabito, I. F. Akyildiz, and M. Johnson, "RCS: A Rate Control Scheme for Real-Time Traffic in Networks with High Bandwidth-Delay Products and High Bit Error Rates," In Proceedings of Infocom 2001, Anchorage, AK, 2001.RCS: A Rate Control Scheme for Real-Time Traffic in Networks with High Bandwidth-Delay Products and High Bit Error RatesRCS: A Rate Control Scheme for Real-Time Traffic in Networks with High Bandwidth-Delay Products and High Bit Error Rates