1. Experimental Measurement of the Capacity for VoIP Traffic in IEEE 802.11 WLANs Authors : Sangho Shin, Henning Schulzrinne [INFOCOM 2007] Reporter : 林緯彥

2. Motivation  Measure the capacity for VoIP traffic in 802.11b test-bed and compare it with the theoretical capacity and simulation results  Identify factors that have been commonly overlooked in past studies but affect experiments and simulations.

3. Outline  Methodology Theoretical capacity for VoIP traffic Theoretical capacity for VoIP traffic VoIP capacity via simulations VoIP capacity via simulations VoIP capacity via experiments VoIP capacity via experiments  “ Hidden factors ” that affect experiments and simulations  Conclusion

4. Methodology  Theoretical capacity for VoIP traffic  Capacity for VoIP traffic via simulation  Capacity for VoIP traffic via experiments  Each method includes two types of VoIP traffic Constant Bit Rate (CBR) Constant Bit Rate (CBR) Variable Bit Rate (VBR) Variable Bit Rate (VBR)

5. Methodology  Theoretical capacity for VoIP traffic  Capacity for VoIP traffic via simulation  Capacity for VoIP traffic via experiments

6. Packetization interval 123N123N ……. MAC Theoretical method parameters value Voice codec 64 kb/s Packet size 160B Packetization interval 20ms Transport layer UDP PHY data rate 11 Mb/s RTS/CTSNo = 15 calls PLCP = Physical Layer Convergence Procedure PLCPMACIPUDPVoiceACKPLCP backoff DIFSSIFS TtTt TbTb RTP Packetization Interval (ms) Capacity (calls)  CBR G.711

7. Theoretical method (cont.)  VBR The VBR VoIP traffic is characterized by talking and silence periods, which determine the activity ratio. The VBR VoIP traffic is characterized by talking and silence periods, which determine the activity ratio. The activity ratio is defined as the ratio of talking-periods at the whole conversation time. The activity ratio is defined as the ratio of talking-periods at the whole conversation time. The activity ratio in the conversation model described in ITU-T P.59 is about 0.39 The activity ratio in the conversation model described in ITU-T P.59 is about 0.39  (CBR capacity / activity ratio) = 15 / 0.39 = 38 calls

8. Methodology  Theoretical capacity for VoIP traffic  Capacity for VoIP traffic via simulation  Capacity for VoIP traffic via experiments

9. Simulation method WIFI Wireless parameters value Voice codec G.711 (64 kb/s) Packet size 160B Packetization interval 20ms Transport layer UDP PHY data rate 11Mb/s RTS/CTSNo WIFI IEEE 802.11b QualNet simulator v3.9 Ethernet

10. Simulation method (cont.)  CBR

11. Simulation method (cont.)  VBR

12. Methodology  Theoretical capacity for VoIP traffic  Capacity for VoIP traffic via simulation  Capacity for VoIP traffic via experiments

13. Experiment method parameters value Voice codec G.711 (64 kb/s) Packet size 160B Packetization interval 20ms Transport layer UDP PHY data rate 11Mb/s RTS/CTSNo client clientsclientAPclient IEEE 802.11b Atheros chipset MadWifi-0.9.3

14. Experiment method (cont.)  CBR

15. Experiment method (cont.)  VBR

16. summary capacityCBRVBR theoretical 15 calls 38 calls simulation 15 calls 34 calls experiment 15 calls 36 calls

17. Factors  Preamble size  Rate control  VoIP packet generation intervals among VoIP sources  Other minor factors

18. Preamble size  Preamble default ： QualNet, NS-2  Long preamble QualNet, NS-2  Long preamble Experiment  Short preamble Experiment  Short preamble  Theoretical capacity with the long preamble = 12 calls LongShort Preamble size 144 us 72 us Header size (us) 48 us 24 us Total size (us) 192 us 96 us Fraction in a VoIP (size) 9%6% Fraction in a VoIP (time) 53%36% long short

19. Rate control  ARF (Auto Rate Fallback) PHY data rate are automatically changes PHY data rate are automatically changes When frame loss is caused by bad link quality, it helps When frame loss is caused by bad link quality, it helps When frame loss is caused by congestion, it makes worse When frame loss is caused by congestion, it makes worse  Problems The effect varies according to algorithms The effect varies according to algorithms Turned off in simulationsTurned off in simulations Turned on in wireless cardsTurned on in wireless cards 8% of frames were transmitted with lower rates

20. Offset of VoIP traffic start time 1234 Packetization interval 1234 Application layer Offset MAC layer databackoff SIFS ACK DIFS data 650 μs = the optimal offset

21. Other minor factors  Scanning APs  Network buffer size and packet loss

22. Scanning APs AP client Probe request (broadcast) Probe response (unicast) Probe request and response frames increase the delay of VoIP packet transmission due to traffic increase

23. Network buffer size and packet loss  Packet loss happens mostly because of the buffer overflow at the AP Small buffer  increase the packet loss Small buffer  increase the packet loss Bigger buffer  reduces packet loss, but increase the delay Bigger buffer  reduces packet loss, but increase the delay

24. Conclusion  Need to consider the following factors when measuring the VoIP capacity experimentally Rate control Rate control Preamble size Preamble size Offset of VoIP traffic start time Offset of VoIP traffic start time  The study can be used in 802.11 experiments and the analysis and comparison

25. Thank you!