Diagnosing Wireless Packet Losses in 802.11: Separating Collision from Weak Signal

1. Diagnosing Wireless Packet Losses in 802.11:Separating Collision from Weak Signal ShravanRayanchu, AruneshMishra, DheerajAgrawal, SharadSaha, SumanBanerjee

2. Motivation • Packet Loss • 2 Causes • Solution Inadequate • 802.11 • Can we determine cause of packet loss?

3. Packet loss in Wireless Networks C A B

4. Packet loss in Wireless Networks A send RTS to B C A B

5. Packet loss in Wireless Networks While A is transmitting, C initiates RTS to B C A B

6. Packet loss in Wireless Networks C A B Since neither A nor B knows the other is transmitting, both RTS’s are sent and collide at B, resulting in packet loss

7. Packet loss in Wireless Networks C A B Since neither A nor B knows the other is transmitting, both RTS’s are sent and collide at B, resulting in packet loss

8. Packet loss in Wireless Networks B C A Here A and C are in just barely in range of each other, but both are in range of B

9. Packet loss in Wireless Networks B C A A send its RTS to C, which is received and B is silenced

10. Packet loss in Wireless Networks B C A C send its CTS to A, but the packet is not heard due to weak signal caused by interference by noise

11. Detecting Packet Loss • Recap: 2 causes of packet loss • 802.11 Solution • BEB • Different causes lead to different solutions

12. Fixing packet loss • Appropriate actions • For collision • BEB

13. Fixing Packet Loss • For low signal • Increase power • Decrease data rate • How to differentiate? Rate = 10 B C E A D Rate = 20

14. Introduction to COLLIE • 802.11, CARA, and RRAA use multiple attempts to deduce cause of packet loss • COLLIE  direct approach • Error packet kickback • Client analysis

15. COLLIE: An Overview • Client Module • AP Module • Server Module (optional)

16. COLLIE: An Overview

17. COLLIE: Single AP • AP error packet kickback • Client-side analysis • Problem: how can the AP successfully re-transmit packet?

18. Experimental Design • Two transmitters, T1 and T2 • Two receivers, R1 and R2 • Receiver R hears all signals

19. Experimental Design • Three possibilities at R: • 1. Packet received without error • 2. Packet received in error • 3. No packet received

20. Error Metrics • Three error metrics: • Bit Error Rates (BER) • Symbol Error Rates (SER) • Error Per Symbol (EPS)

21. Metrics for Analysis • Received Signal Strength (RSS) = S + I • High RSS  collision • Low RSS  channel fluctuations • Bit Error Rate (BER) = total # incorrect bits • BER is higher for collisions, lower for low signal

22. RSS: The Details • Of all packets lost due to low signal, 95% had an RSS less than -73dB, compared to only 10% for collisions

23. Metrics for Analysis • Symbol level errors: errors within transmission frame • Multiple tools used to analyze symbol-level errors

24. Framing • 0011 0011 0011  0011 1101 0011 Collision • 0011 0011 0011  0111 1011 0010 Channel Fluctuation

25. Symbol-level Errors • Symbol Error Rate (SER)- # symbols received in error • Errors Per Symbol (EPS)- average # errors within each symbol • Symbol Error Score (S-score): calculated as , where Bi is a burst of n bits • 74% accuracy

26. S-Score • 0011 0011 0011  0011 1101 0011 Collision • 0011 0011 0011  0111 1011 0010 S-Score = S-Score = Channel Fluctuation

27. Performance • Metric voting scheme • Successful almost 60%, false positive rate 2.4%

28. Some Problems • RSS: universal cutoff impossible • Capture Effect • Packet size

29. Multi-AP COLLIE • Error packet sent to a central COLLIE server • Most important where the capture effect is dominant

30. Results • Static situation  average of 30% gains in throughput • For multiple collision sources and high mobility, throughput gains of 15-60%

31. Conclusions • COLLIE implementation achieves increased throughput (20-60%) while optimizing channel use • Implementation can be done over standard 802.11, resulting in much lower startup costs than other protocols

32. Questions?