VoIP over 802.11 Wireless LAN

Brandon Wilson

PI: Alexander L. Wijesinha

• Linksys Cable/DSL Router w/ an 8-port switch

• Cisco Aironet 1200 series Access Point

• 4 Linksys Wireless-G USB adapters (Soft phones on Windows)

• 3-Com Wireless PCI adapters on Linux PCs

Network configuration

• Obtain a basic understanding of the 802.11 protocol

• Examine the bandwidth consumption of one VoIP call (wireless to wireless)

• Research the effectiveness of the back-off algorithm (in the 802.11 standard)

• Look closer at the tradeoff between call quality and background traffic throughput

Research Objectives

Making a Call

Beacons

• Frequency 100ms

• Packet Size = 240 bytes

• Beacons/sec= 1/.1 = 10/sec

• (10*240 bytes*8 bits)/106= 0.0192 Mbit / sec

1 Call Theoretical Bandwidth

• 50 Packets/sec Voice Data G.711

• Total bytes to transmit all voice packets = (100 * 380) + (102 * 158) + 221 + 226=54563 bytes/sec

• Theoretical Bandwidth of one call = (54563*8) / 106 = 0.436504 Mbit/sec

• FTP Server on the wired LAN and a FTP client on the wireless LAN

• Measure bandwidth as only TCP packets and corresponding 802.11 acknowledgement packets

TCP Traffic

TCP Alone

TCP with one call

TCP with two calls

Avg TCP Bandwidth Utilized with constant TCP traffic and VoIP calls

9.806

8.957

8.104

y = -0.851x + 10.658

0.000

2.000

4.000

6.000

8.000

10.000

12.000

0 1 2

No. Of VoIP Calls

Av

g. T

CP

Ba

nd

wid

th U

tiliz

ed

(1

0 t

ria

l me

an

)

Average TCP Bandwidth

UDP Traffic

• UDP packets generated on wireless LAN by PackETH

• With each added call, adjust UDP packet delay to retain call quality

• Measure Bandwidth as only UDP packets and their corresponding 802.11 acknowledgements (ignore all other traffic)

UDP Only

UDP with one call

UDP with 2 calls

Average UDP Bandwidth

Avg. UDP Bandwidth

8.2582425

9.7880425

10.766141

0

2

4

6

8

10

0 1 2

No. VoIP Calls

Ba

nd

wid

th (

Mb

it/s

ec

)

Conclusions

TCP• The results of the experiments showed that TCP

traffic degrades to a significant extent when a VoIP call is added to the network

• A possible explanation for the decrease is the TCP congestion control mechanism or the access point may be giving priority to the VoIP traffic

• ~ 0.85 Mbit / sec drop in TCP Bandwidth which is very significant compared to ~ 0.45 Mbit / sec of bandwidth require by a VoIP call

Conclusions cont’d

UDP

Losses in bandwidth are far greater than the average

bandwidth used by a VoIP call and show that the call,

with assured voice quality, consumes a much greater

portion of the 802.11 bandwidth than expected

1 call ~ 1 Mbit / sec bandwidth loss

2 call ~ 1.5 Mbit / sec bandwidth loss

Future Work

Future work will consist of repeating the same experiments on a 802.11g network and looking at the effects and behavior of the network when the only variation is the bandwidth is increased by about a factor of 5