Deployment Guidelines for Highly Congested IEEE 802.11b/g Networks

Andrea G. Forte and Henning Schulzrinne

Columbia University

Background

• IEEE 802.11b/g networks widely spread• Mostly uncoordinated deployments• Channel assignment not trivial

• Our approach– Test on the field (enough optimal algorithms!)– Come out with practical guidelines for network

administrators

Wireless Networks

• IEEE 802.11a– 5 GHz band– 12 non-overlapping channels– Scarcely deployed and used

• IEEE 802.11b/g– 2.4 GHz band– 3 non-overlapping channels– Widely spread

We focus on IEEE 802.11b/g IEEE 802.11b

Problems (1/3)Handoff behavior (65th IETF meeting)

• Handoff is triggered – generally, by low signal

strength – in congested channel, by

frame loss

• Effect of layer 2 handoff– Increase of traffic– Disruption of network

(0.5 ~ 1.5 sec)

306

222

84

111

112

36

262

227

162

218

183

131

0

100

200

300

400

500

600

700

Nu

mb

er

of

ha

nd

off

sS1 Lunch S1 P

IETF Sessions

C11

C6

C1

The number of handoff per hour in each IETF session

Problems (2/3)Handoff behavior (65th IETF meeting)

1→122%

1→69%

1→113%

6→19%

6→633%

6→112%

11→13%

11→62%

11→1117%

• Handoff to the same channels : 72%

• Handoff to the same AP : 55%

• Handoffs between channels

Problems (3/3)Handoff behavior (65th IETF meeting)

< 5 min33%< 10 min

12%

< 1 min23%>= 10 min

32%

• Too often handoff– Disruption of network

• 0.5 ~1.5 sec per handoff

– Increase of traffic due to handoff related frames – probe request and response

– 10.4% of total

• Distribution of session time: time between handoffs

Experiments

• Experiment 1– Testing different channel configurations in existing

networks• Columbia University campus (site survey)

• Experiment 2– Studying co-channel interference in highly

congested scenarios (large number of users)• ORBIT wireless test-bed

Site Survey – Columbia University

Google Map!

Site Survey – Columbia University

• Found a total of 668 APs– 338 open APs: 49% – 350 secure APs: 51% – Best signal: -54 dBm– Worst signal: -98 dBm

• Found 365 unique wireless networks– “private” wireless networks (single AP): 340– “public” networks (not necessarily open): 25

• Columbia University: 143 APs• PubWiFi (Teachers College): 33 APs• COWSECURE: 12 APs• Columbia University – Law: 11 APs• Barnard College: 10 APs

Experiment 1Experimental setup

AP Client

Sniffer

Surrounding APs Surrounding APs

Experiment 1 – Results (1/3)Using non-overlapping Channels

0

10

20

30

40

50

60

70

80

1 2 3 4 5 6 7 8 9 10

Experiment

Kb

/s

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

%

Tput

Retries

0

10

20

30

40

50

60

70

80

1 2 3 4 5 6 7 8 9 10

Experiment

Kb

/s

0

10

20

30

40

50

60

70

%

Tput

Retries

•Throughput and retry rate with no interference

Same for any channel

•Throughput and retry rate with interference on channel 1

Experiment 1 – Results (2/3)Using non-overlapping Channels

0

10

20

30

40

50

60

70

80

1 2 3 4 5 6 7 8 9 10

Experiment

Kb

/s

0

10

20

30

40

50

60

70

%

Tput

Retries

0

10

20

30

40

50

60

70

80

1 2 3 4 5 6 7 8 9 10

Experiment

Kb

/s

0

10

20

30

40

50

60

70

%

Tput

Retries

• Throughput and retry rate with interference on channel 6

Most congested!

•Throughput and retry rate with interference on channel 11

Experiment 1 – Results (3/3)Using Overlapping Channels

0

10

20

30

40

50

60

70

80

1 2 3 4 5 6 7 8 9 10

Experiment

Kb

/s

0

10

20

30

40

50

60

70

%

Tput

Retries

0

10

20

30

40

50

60

70

80

1 2 3 4 5 6 7 8 9 10

Experiment

Kb

/s

0

10

20

30

40

50

60

70

%

Tput

Retries

• Throughput and retry rate with interference on channel 4

Better than channel 6

• Throughput and retry rate with interference on channel 8

Better than channel 6

Experiment 1Conclusions

• Using overlapping channels does not affect performance negatively– In the experiments channel 4 and channel 8 are a

much better choice than channel 6

• Use at least channels 1, 4, 8 and 11 (minimum overlapping in band) – better spatial re-use – no significant decrease in performance

USE OVERLAPPING CHANNELS!

Experiment 2Experimental setup

• ORBIT wireless test-bed– Grid of 20x20 wireless nodes– Used only maximum bit-rate of 11 Mb/s (no ARF)– G.711 CBR– Number of clients always exceeding the network

capacity (CBR @ 11Mb/s 10 concurrent calls)

Experiments 2 – Results (1/2)Non-overlapping Channels

0

10

20

30

40

50

60

1 2 3 4 5 6 7 8 9 10

Experiment

%

0

0.5

1

1.5

2

2.5

Mb

/s

Tput

PHY-err

CRC-err

long ret

0

10

20

30

40

50

60

1 2 3 4 5 6 7 8 9 10

Experiment

%

0

0.5

1

1.5

2

2.5M

b/s

Tput

PHY-err

CRC-err

long ret

• AP1 using Ch. 1

• AP2 using Ch.6

• Num. of clients: 43

• AP1 and AP2 using Ch. 1

• Num. of clients: 43

Experiments 2 – Results (2/2)Overlapping Channels

0

10

20

30

40

50

60

1 2 3 4 5 6 7 8 9 10

Experiment

%

0

0.5

1

1.5

2

2.5

Mb

/s

Tput

PHY-err

CRC-err

long ret

0

10

20

30

40

50

60

1 2 3 4 5 6 7 8 9

Experiment

%

0

0.5

1

1.5

2

2.5M

b/s

Tput

PHY-err

CRC-err

long ret

• AP1 using Ch. 1

• AP2 using Ch. 4

• Num. of clients: 67

• AP1 and AP2 using Ch. 4

• Num. of clients: 67

Experiment 2Conclusions

• When using two APs on the same channel– Throughput decreases drastically– Physical-error rate and retry rate increase

• Using two APs on two overlapping channels performs much better than using the same non-overlapping channel

Do not deploy multiple APs on the same non-overlapping channels

USE OVERLAPPING CHANNELS!

One AP vs. manyVery high number of users

0

10

20

30

40

50

60

1 2 3 4 5 6 7 8 9 10

Experiment

%

0

0.5

1

1.5

2

2.5

Mb

/s

Tput

PHY-err

CRC-err

long ret

0

10

20

30

40

50

60

1 2 3 4 5 6 7 8 9 10

Experiment

%

0

0.5

1

1.5

2

2.5M

b/s

Tput

PHY-err

CRC-err

long ret

•Network performance with single AP in highly congested scenarios

•Network performance with two APs on the same channel in highly congested scenario

Using two APs on the same channel performs worst than using a single AP!

Conclusions

• Using overlapping channels does not affect performance negatively– Use at least channels 1, 4, 8 and 11

• Do not deploy multiple APs on the same non-overlapping channels

• Using two APs on the same channel performs worst than using a single AP!– Just increasing the number of APs does not help

USE OVERLAPPING CHANNELS!