Dynamic channel allocation
in wireless ad-hoc networks
Anup Tapadia
Liang Chen
Shaan Mahbubani
Background
802.11 networks have experienced a
tremendous growth in last few years
 Access points work on a single static
channel
 Contention & congestion
 Low throughput due to inefficient channel
allocation
 Presence of other interfering networks
impacts performance

Channel variations
Traffic on Thursday 26th April at Calit2 6th floor
90000
Channel 1
Channel 6
Channel 11
Packet Count
80000
70000
60000
50000
40000
30000
20000
10000
0
1
3
5
7
9 11 13 15 17 19 21 23
Hour of the day
CogNet testbed setup at Calit2
Cognet Database
Channel Controller
Dynamic Link
Monitor 1 Node 1
Node 2 Monitor 2
CogNet is supported by
NSF SGER-CogNet
CalNode




Seokris 4521
Linux 2.6
Iperf
Monitor Card


Data Card


Atheros w/ Mad Wifi
Prism 2.5 802.11b
Wired to UCSD
network
What may affect the channel
performance?

Interfering 802.11 Traffic
Packet count
 802.11 Rate
 RSSI


Other interference
Bluetooth
 Cordless phones
 Microwaves

Our Focus
Challenge

How do we select the channel that
permits maximum throughput in an
ad-hoc network ?
Channel Controller (Java & C++)
CogNet
DB
SQL Query
Decision Logic
AP
Connection
Stub
iwlib
WiFi WiFi
NIC NIC
Towards Nodes
Channel
Switcher
Iperf Test
Scheduler
Software Architecture
Channel Controller
Decision Logic
Channel Switcher
AP
AP
AP
AP
Connection
Stub
Connection
Stub
Connection
Stub
Connection
Stub
iwlib
iwlib
iwlib
iwlib
WiFi WiFi
NIC NIC
WiFi WiFi
NIC NIC
WiFi WiFi
NIC NIC
WiFi WiFi
NIC NIC
Channel Switch
Channel Controller
Decision Logic
Channel Switcher
Channel Switch Message
Switch
Channel
Channel Switch Message
Connection
Stub
Connection
Stub
iwlib
iwlib
802.11
Driver
802.11
Driver
Probe
802.11 H/W
Reply
802.11 H/W
Switch
Channel
Factors selection test
Tx
Tx
15 sec silence
Tx
15 sec silence
Factors affecting throughput :RSSI
Effect of our transmission rate on
others RSSI
60
Our packets
Other packets in presence of our packets
Packets during silence period
55
50
Average RSSI / Packet
45
40
35
30
25
20
15
10
0
1000
2000
3000
4000
Applied Rate (Kbps)
5000
6000
7000
Factors affecting throughput : Rate
Factors affecting throughput :Packets
Channel Selection Algorithm

We select the least interference channel
dynamically

Channel interference metric
N ip  N iq
Ci 
2
where Ci is the interference metric of channel i, Nip, Niq
are the number of other packets captured at node p
and q on channel i in a 120s window.
 Selected channel j, s.t.
C j  min (Ci )
i 1, 6,11
Experiment: Process
120 sec query window
S
R
15 sec silence



D
15 sec silence
Send a 3MB file using each scheme
15s pause interval in between
each transfer
Repeat
Background Traffic
6000
Other packets count
5000
4000
Ch1PktCount
3000
Ch6PktCount
Ch11PktCount
2000
1000
0
0
50
100
Run # (Progressive time)
150
200
Results: Dynamic v. Static
Results: Dynamic v. Random
Results: Random v. Static
Number of times channel was selected
Channels picked
70
Dynamic
60
Static
Random
50
40
30
20
10
0
1
6
11
Channel #
Results: Summary
Scheme
Dynamic
Static
Random
Avg Rate
4042 kbps
2100 kbps
2742 kbps
Average
Transfer
Time (3MB)
6.08s
11.7s
8.96s
Effect of channel switching time
t d  tc  t s
td  t s  tc
f
t d   tc
rs
rd 
for tc  5.7 s
f  3072kbits
rs  2100kbps
rd  4096kbps
f
f
 tc
rs
td = Tx time on dynamic channel
ts = Tx time on static channel
tc = Time to change
rd = Rate on dynamic channel
rs = Rate on static channel
f = Size of file
Our implementation
Future work
Multi hop links
 More complex allocation schemes
 Finer data granularity and patterns
 Improve channel switching time
 Effect of frequency of channel switching

Conclusions
Dynamic performed better than Static and
Random
 Static on a free channel provides upper
bound on Dynamic performance;
improvements come when static channel
is loaded (channel usage varies over time)
