Professional Documents
Culture Documents
5, MAY 2014
2770
I. I NTRODUCTION
Wireless LANs (WLAN) provide quick access to the Internet and are widely deployed in hotspot areas. Typically all
nodes in a WLAN access the Internet by the associated Access
Point (AP) and share the common wireless channel with the
well-known Carrier Sense Multi-Access (CSMA) scheme [1].
CSMA has achieved great success in the Ethernet where collision can be detected quickly. In wireless networks, however,
due to the half-duplex nature, each node cannot simultaneously
transmit and receive with the same radio. Instead collision
avoidance is used by the virtual carrier sense mechanism
[2], [3]. Lack of a timely ACK after the transmission of a
data packet is regarded as a collision. Then exponential backoff is started to alleviate collision. CSMA works fairly well
in times of light traffic. However, as the number of nodes
in a WLAN increases quickly, severe collision unavoidably
degrades performance of the whole network. Since quick
detection of collision is infeasible in a WLAN, the only choice
to reduce collision is to control the number of nodes attending
the contention. Some conventional schemes include channel
diversity [4], directional antenna, transmit power control,
adaptive backoff [5] and carrier sense [6]. Their common idea
is to reduce the number of nodes in a collision domain.
Multi-path fading degrades performance of a WLAN with
random signal fluctuation. It is usually regarded as being harmful and different diversity schemes, such as receive diversity,
transmit diversity and time diversity, have been proposed to
mitigate fading. Retransmission is also an effective solution.
Though either fading or increase in node density affects
system performance, a joint consideration of the two factors
may even improve system performance by exploiting multiuser scheduling ([7], Chapter 6). The basic idea is to exploit
the independence of random fading. In cellular networks an
AP can monitor Channel State Information (CSI) all the time
and perform centralized scheduling. Multi-user diversity has
already become a part of 3G wireless communication systems.
In a WLAN, however, there is no feedback channel and an AP
cannot detect CSI if a node keeps silence.
In this paper we propose a Distributed Multi-User Scheduling (DMUS) scheme, aiming to improve scalability and
throughput of WLANs. Time-variant link quality under multipath fading is considered and nodes are allowed to transmit
only when their link quality is high enough. By setting a
suitable SNR threshold, at any instance only a small percentage of nodes contend to access the common wireless
channel. This SNR threshold is set by the AP, according to
the number of nodes in the cell. In addition, it is a normalized
value, which enables fair access of the channel by all nodes.
In conventional CSMA a node has to double its contention
window and perform exponential back-off in times of collision.
In the proposed DMUS scheme a fixed contention window
is adopted and the collision probability is controlled by the
AP. The total throughput of a WLAN is improved since
transmissions are finished at higher rate by avoiding deep
fading. Simulation results indicate that DMUS can improve
throughput by up to 194.6% compared with CSMA/CA in
Rayleigh fading environment when there are 40 nodes in a
WLAN.
The rest of the paper is organized as follows. Section II
reviews previous efforts on improving throughput of WLANs.
Then research motivation is addressed in Section III. In
Section IV we propose our system model and discuss protocol
design. In Section V by analysis we show how to find
optimal parameters for minimizing collision and achieving
high throughput. Simulation results and analysis are presented
in Section VI. Finally we conclude the paper with Section VII.
II. BACKGROUND AND R ELATED W ORK
CSMA in a WLAN depends on distributed contention and
does not scale well to node density. As a result an increase
in the number of nodes in a WLAN leads to higher collision
probability and degrades network throughput. Previous efforts
have suggested quick collision resolution or adaptive carrier
sense. Others even suggest benefiting from the increase of
nodes by exploiting multi-user diversity.
2771
???
????
2772
SNR
C
AP
B
A
AP Poll
SNR
ACK
DIFS
Poll
N*SLOT
PBU
SIFS
SLOT
SNR
ACK
DIFS
PAU
Fig. 1.
SNR
PCU
Fig. 2.
SNR
SNR
AP Poll
A
B
SNR
PAD
DIFS
CTS
SLOT
SIFS
SNR
PBD
Poll
ACK
N*SLOT
CTS
SIFS
ACK
DIFS
Fig. 3.
2773
m = 1,
1,
N
1
P
Pm (0 ) =
(N i)m1 /N m , m 2.
m
(2)
i=1
(4)
probability is
(1)
(6)
where trand is the random slotted waiting time before transmission, L is the packet length and r() is the actual rate.
tSIF S is the duration of SIFS and tACK is the transmission
time of the ACK frame.
(iii) More than two links have normalized SNR greater than
0 and a collision occurs. The probability is
PC (0 ) =
M
X
T
Pm
(0 )(1 Pm (0 )) = 1 PE (0 ) PS (0 ),
m=2
(7)
(8)
PS (0 )tS
.
PE (0 )tE + PS (0 )tS + PC (0 )tC
(9)
(10)
(11)
2774
Probability PS (0 )
1
0.8
0.6
NumOfSlots
NumOfSlots
NumOfSlots
NumOfSlots
NumOfSlots
0.4
0.2
=
=
=
=
=
1
5
10
20
50
0
0
Fig. 4.
Probability PS (0 )
1
0.8
0.6
0.4
0.2
NumOfNodes
NumOfNodes
NumOfNodes
NumOfNodes
=
=
=
=
2
5
15
30
0
0
Fig. 5.
2775
Channel utility
0.8
0.7
CSMA/CA
ContentionFree
DMUS
0.6
0.5
0.4
10
20
30
40
Fig. 6.
ACKNOWLEDGMENT
This work was supported by National Institute of Information and Communications Technology (NICT) of Japan.
Throughput (Mbps)
R EFERENCES
2
CSMA/CA
ContentionFree
DMUS
0
10
20
30
40
Fairness
0.99
0.98
CSMA/CA
ContentionFree
DMUS
0.97
10
20
30
40
Fig. 8.