The Impact of Channel Usage Information on
the Throughput Achieved by 802.11-Style
MACs in Urban Mesh Networks
Jonghyun Kim
Stephan Bohacek
Department of Electrical and Computer Engineering
University of Delaware
Outline
Introduction
Throughput Metric
Date-Rate Selection
IEEE 802.11 Variants
Idealized 802.11-style MACs
Simulation Environment
Simulation Results
Conclusions
Introduction
•
Why does optimal spatial TDMA (STDMA) achieve
higher throughput than CSMA/CA such as IEEE
802.11?
 By eliminating collisions
 By ordering packets in an optimal way
Introduction
A
T1
C
W
X
GW
Y
Z
B
D
A
C
T2
W
X
GW
Y
Wait 1
Z
B
D
A
C
T3
W
X
GW
Y
Wait 2
Z
B
D
A
C
T4
W
B
X
GW
Y
Destination : A
Destination : B
(same side)
Z
D
Introduction
A
T1
C
W
X
GW
Y
Z
B
D
A
C
T2
W
X
GW
Y
Wait 1
Z
B
D
A
C
T3
W
X
B
GW can transmit 50% sooner
GW
Y
Destination : A
Destination : C
(opposite side)
Z
D
Introduction
 Goal
•
•
Investigate how to increase the throughput by various
802.11-style MACs with different techniques and extending
channel usage information so that the throughput may
approach gradually and maximally to the throughput
achieved by optimal STDMA.
Investigate the main reasons (with percentage information)
that optimal STDMA achieves higher throughput than
802.11
Introduction
 Topology
: Gateway
: Destination (mesh node)
Downtown Chicago (2kmx2Km)
3 Gateways with18 destinations
Outline
Introduction
Throughput Metric
Date-Rate Selection
IEEE 802.11 Variants
Idealized 802.11-style MACs
Simulation Environment
Simulation Results
Conclusions
Throughput Metric
500
450
2
F bps
1
4
flow1
3
5
6
g(F) (Kbps)
400
350
Throughput
Peak
300
250
flow1
flow2
flow3
flow4
flow5
flow6
200
150
100
200
250
300
350
400
450
F (Kbps)
F
Transmission rate to each destination 
g (F )
Average arrival rate at destination 
min g ( F )
Minimum average arrival rate
max min g ( F )
Peak value (Throughput)

F

500
500
Outline
Introduction
Throughput Metric
Date-Rate Selection
IEEE 802.11 Variants
Idealized 802.11-style MACs
Simulation Environment
Simulation Results
Conclusions
SNR1
Link 1
Effective data rate (Mbps)
Data-Rate Selection
40
Peak
35
SNR = 18 dB
30
25
Data-rate
20
15
10
5
0
0
10
20
30
40
50
60
R (Mbps)
R
Bit-rate provided by 802.11a (6,9,12,…,54 Mbps)
PSP( SNRi , R)
Prob. of successfully decoding a packet
transmitted over link i at bit-rate R
max R  PSP( SNRi  Guard , R)
Maximum effective data rate over link i
arg max R  PSP( SNRi  Guard , R)
Data-rate for link i
R
R
Outline
Introduction
Throughput Metric
Date-Rate Selection
IEEE 802.11 Variants
Idealized 802.11-style MACs
Simulation Environment
Simulation Results
Conclusions
IEEE 802.11 Variants
 Standard 802.11
overhearing
S1
S2
DATA
DATA
R1
overhearing
ACK
R2
Node S1
Layer
MAC
State
DIFS
NAV
BO
DIFS BO remaining
time
DATA frame from S2
ACK frame from R2
Receive
PHY
T1
DATA frame to R1
time
Transmit
time
DIFS
BO
Busy
DIFS BO
Transmit
IEEE 802.11 Variants
 Stomp
S1
S2
DATA
R1
R2
Layer
MAC
ACK
State
DIFS
BO
time
DATA frame from S2
Receive
time
PHY
Decode PLCP header
Transmit
DIFS
time
BO
IEEE 802.11 Variants
 Stomp
SNRS1-S2
S1
S2
SNRS1-R2
SNRS1-R1
Data-rate
SNRR1-R2
Layer
MAC
SNRS2-R2
Data-rate
SNRR1-S2
R1
Red characters : channel usage information S1 knows
R2
State
DIFS
BO
BO remaining
time
DATA frame from S2
Receive
time
PHY
Decode PLCP header
DATA frame to R1
Transmit
DIFS
time
BO
Transmit
IEEE 802.11 Variants
 Packet Reordering
Buffer
B
S1
S2
overhearing
L3
L1
A
Packet (destination = A)
Packet (destination= B)
L2
Link 1
Link 2
Conflict
Link 3
Link 2
No conflict
DATA
R2
IEEE 802.11 Variants
 Capture of Stronger Signals
S1
R
S2
R
DATA frame from S2
Receive
time
PHY
T1
DATA frame from S1
Receive
time
T2
Received power
PS1  PS2
PS2
PS1
time
Outline
Introduction
Throughput Metric
Date-Rate Selection
IEEE 802.11 Variants
Idealized 802.11-style MACs
Simulation Environment
Simulation Results
Conclusions
Idealized 802.11-Style MACs
 Transmitter-Transmitter Regional Channel Usage Information (TTCUI)
T5
R5
Transmitting nodes :
T2
36Mbps
6Mbps
regionregion
R3
T3
R6
T6
T2
T3
T4
T5
R2
T1
T6
R1
R4
T4
Idealized version of 802.11 without RTS/CTS
Idealized 802.11-Style MACs
 Transmitter-Transmitter Regional Channel Usage Information (TTCUI)
36Mbps region
T1
R1
Cause a collision
R4
T4
Idealized 802.11-Style MACs
 Transmitter Regional Channel Usage Information (TCUI)
T2
36Mbps region
R3
T3
R2
T1
Transmitting nodes :
T2
T3
Receiving nodes :
R2
R4
R1
R4
T4
Half idealized version of 802.11 with RTS/CTS
Idealized 802.11-Style MACs
 Transmitter Regional Channel Usage Information (TCUI)
36Mbps region
Cause a collision
T1
R1
T5
R5
Idealized 802.11-Style MACs
 Transmitter and Receiver Regional Channel Usage Information (TRCUI)
T6
T2
R2
R6
T1
R1
T5
R5
R7
R3
T3
R4
T4
T7
Transmitting nodes :
T2
T3
Receiving nodes :
R2
R4
T4
T5
Idealized version of 802.11 with RTS/CTS
Idealized 802.11-Style MACs
 Global Channel Usage Information
T6
T2
R2
R6
T1
R1
T5
R5
R7
R3
T3
R4
T4
T7
Transmitting nodes :
T2
T3
T4
T5
T6
T7
Receiving nodes :
R2
R3
R4
R5
R6
R7
Outline
Introduction
Throughput Metric
Date-Rate Selection
IEEE 802.11 Variants
Idealized 802.11-style MACs
Simulation Environment
Simulation Results
Conclusions
Simulation Environment
 Simulation Set-up
# of gateways
1, 3, 5
# of destinations
18, 36, 54, 72, 90
# of topology samples
10
# of topologies
150
City map
Downtown Chicago (2Km x 2Km)
Size of topology
6 x 6 city block randomly chosen from city map
Application traffic
CBR with 1344 bytes and F bps
MAC protocol
802.11a with transmission power of 18 dBm
Mobility
UDel Mobility Model
Channel gain
UDel Channel Model
Packet simulator
QualNet
Methodology used to
determine the throughput
Golden section method and bootstrap percentile
confidence interval
Simulation Environment
 MAC types
Type
802.11-style MAC algorithm
T2
T2R2
R2
A
Without RTS/CTS (baseline)
B
With RTS/CTS transmitted at 6Mbps
C
With CTS-to-self transmitted at 6MbpsT1
D
Aloha-like
E
R4
Without RTS/CTSR3and captureT3R3
of stronger T3
signals
F
Without RTS/CTS and with stomp, packet reordering,
capture of stronger signals
G
TTCUI with 6Mbps region
H
TCUI with 6Mbps region
I
TRCUI with 6Mbps region
J
Global knowledge
K
Optimal STDMA
6Mbps region 6Mbps region
T1
R1
Stomp, packet reordering,
capture of stronger signals
T5
R1
T4
R4
R5
T4
Idealized
Outline
Introduction
Throughput Metric
Date-Rate Selection
IEEE 802.11 Variants
Idealized 802.11-style MACs
Simulation Environment
Simulation Results
Conclusions
Simulation Results
# of gateways = 1
Ratio of throughput
3.5
# of gateways = 3
# of gateways = 5
3.5
3.5
3
3
2.5
2.5
2.5
2
2
2
1.5
1.5
1.5
1
1
1
3
0.5
20
40
60
80
100
0.5
20
40
60
80
100
0.5
20
40
60
80
100
# of destinations
A (Without RTS/CTS) – baseline
T2
R2
T2
R2
B (With RTS/CTS)
C (With CTS-to-self )
D (Aloha)
E (Without RTS/CTS and capture of stronger signal)
T1
T1
R1
T5
R1
F (Without RTS/CTS and with stomp, packet reordering, capture of stronger signal )
G (TTCUI)
H (TCUI)
R3
R4
R3
T3
R4
T4
T3
I&J (TRCUI & GK)
K (Optimal STDMA)
R5
T4
Outline
Introduction
Throughput Metric
Date-Rate Selection
IEEE 802.11 Variants
Idealized 802.11-style MACs
Simulation Environment
Simulation Results
Conclusions
Conclusion
•
•
•
•
Significant improvement does not occur with perfect
knowledge of only nearby transmitters.
Significant improvement starts to occur with perfect
knowledge of nearby transmitters and receivers.
Half collisions will occur with knowing perfect channel
activity around only transmitter.
Half performance improvement by optimal STDMA is
because of elimination of collisions. The other is
because of ordering packets in an optimal way.
Thanks
Any questions, comments, suggestions ?
E-mail : [email protected]
[email protected]
UDel Models – Website
http://udelmodels.eecis.udel.edu
Extra Slides
0
Prob. of bit error
10
-1
10
-2
10
-3
10
-4
10
-15
6Mbps
9Mbps
12Mbps
18Mbps
24Mbps
36Mbps
48Mbps
54Mbps
-10
-5
0
5
SNR (dB)
10
15
20
Extra Slides
Frame in 802.11a PHY
CBR data size = 1344 bytes
UDP header size = 8
IP header size = 20
DATA MAC header size = 28
MSDU size = 1372
PSDU size = 1400 bytes
SIFS = 16 us
DIFS = 34 us
EIFS = 94 us
SLOT_TIME = 9 us
CW_MIN = 15 (contention window)
CW_MAX = 1023
Question :
How much collision occur in this scheme?
For DATA
PLCP preamble
Signal
Service
MAC Hdr
MSDU
Tail
96 bits
24 bits
16 bits
28 bytes
1372 bytes
6 bits
Duration = 20 us
Duration = 4.44 us
Bit rate = 6 Mbps (fixed)
Bit rate = 54 Mbps (varies)
Duration = 203.37 us
Our objective is to use the channel during this time if conditions are satisfied based
on sender, receiver, and duration from DATA MAC header.
Conditions
1. If receiver is not the node that is now receiving the frame above.
2. If the link between sender and receiver is not a carrier sensing neighbor of the link between this node
and the node’s receiver which is determined later if MAC does not have a frame received from network
layer now or it already have a frame, but it stops counting back-off due to channel busy. Need to adjust
back-off time.
Pad
254
This is the current
last point
# of trials are increased
252
in four times
Actual bit rate (Kbps)
250
Tracking line
(B point)
248
246
244
Tracking line
(A points)
242
240
Confidence Interval
238
236
265
270
275
280 285 290 295 300
Desired bit rate (Kbps)
305
310
315
Golden Section Method and bootstrap to find confidence
interval are used to find optimal capacity (i.e., bit rate) in
urban mesh network.
T2
6Mbps region
R3
T3
R2
T1
T2
6Mbps region
R1
R4
T4
R3
T3
R2
T1
R1
R4
T4
T2
6Mbps region
R3
T3
R2
T1
R1
R4
T5
T4
R5