Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

2013 4th International Conference on Intelligent Systems, Modelling and Simulation

C-OFDMA: Improved Throughput for Next Generation WLAN Systems Based on


OFDMA and CSMA/CA

Gebreselassie Haile Jaesung Lim


Tactical Network Research Center Tactical Network Research Center
Ajou University Ajou University
Suwon, South Korea Suwon, South Korea
gebsela@gmail.com {jklee64, jaslim}@ajou.ac.kr

Abstract— In Orthogonal Frequency Division Multiple based on a combination of OFDM and TDMA to provide
Access (OFDMA), there are multiple sub-channels which better performance through intelligent assignment of
enable multiple stations to transmit concurrently on different subcarriers in time domain, but this scheme is not designed
sub-channels. The stations can obtain the channel information for efficient random access. Other authors in [3] [4] studied
of all the sub-channels at each time instant. Carrier Sense dynamic OFDMA in IEEE 802.11 systems where extended
Multiple Access with Collision Avoidance (CSMA/CA) allows control messages based on information from the Access
different stations to randomly contend for the available sub- Point (AP) are used to solve sub-channel allocation
channels from the time domain to avoid collision with each problems. In [5], OFDMA with ALOHA random access
other. This gives a resource in time and frequency domain. In
scheme is proposed which avoids collisions by changing
this paper, we present a Concurrent OFDMA-based
CSMA/CA (C-OFDMA) MAC protocol for Next Generation
sub-channels. This scheme doesn’t use carrier-sensing, has
Wireless Local Area Network where all the stations that high collision probability among the terminals, It also has
transmit successful Request-To-Send (RTS) packet in the complex scheduling, and therefore results in low throughput.
contention phase send their traffic concurrently to the Access The authors in [6] [7] proposed multi-channel random access
Point (AP). The proposed C-OFDMA protocol is compared scheme for OFDMA-based CSMA/CA where the back off
with existing IEEE 802.11 DCF, IEEE 802.11 RTS/CTS, and counter of each STA decreases by the number of idle sub-
Hybrid-OFDMA MAC protocols and simulation results show channels at each time instant, and transmits whenever it
that it provides improved throughput for crowded WLAN reaches zero. These protocols provide better throughput than
systems. the multi-channel ALOHA protocol [5], but they don’t use
four-way hand-shaking scheme, known as Request-To-
Keywords—OFDMA, Sub-channel, CSMA/CA, WLAN, Send/Clear-To-Send (RTS/CTS), which can be used to
Concurrent, Interframe Space detect collisions easily during the random access. Also, there
is no clear distinction between the data and
I. INTRODUCTION acknowledgement (ACK) packet transmissions.
As the number of Stations (STAs) in wireless systems In [8] [9], the authors proposed a Hybrid MAC protocol
increases, the throughput highly depends on the Medium based on OFDMA and CSMA/CA (H-OFDMA) to increase
Access Control (MAC) protocol and its associated Multiple throughput in next-generation WLAN systems. It combines
Access. One of the promising multiple access approaches in OFDMA access with a CSMA/CA scheme and uses a two-
wireless networks is Orthogonal Frequency Division stage frame delivery process: Transmission Opportunity
Multiple access (OFDMA). OFDMA has become an Request (TR) and Scheduled Transmission (ST). Request-
important multiple access technology for key next- To-Send (RTS) packet is randomly sent by applying
generation wireless networks such as WiMAX [1]. One of OFDMA-based CSMA/CA scheme in the TR phase. Fig. 1
the advantages of OFDMA is that the whole frequency shows the frame structure of the H-OFDMA MAC protocol.
spectrum can be divided into multiple narrow-band sub- When the number of sub-channels (SCHs) is smaller than the
channels and the terminals can be assigned different sub- number of STAs, more than one STA has to contend for each
channels to transmit their traffic simultaneously according to SCH which means that there is probability of collision. A
a certain MAC mechanism. There are two types of MAC CSMA/CA backoff scheme is used to resolve collisions.
mechanisms that can be adopted for wireless networks: Once the AP receives the RTS packets, transmission
random access and scheduled access. Scheduled access has schedule for the each STA that received a CTS packet is
high signaling overhead, and is well suited for connection- broadcast, and each STA transmits their data sequentially
oriented applications that contain clear specifications of using all the subcarriers (OFDM with time domain
traffic demands. For burst data traffic, random access is a multiplexing) in the ST phase. Assuming that there are k
preferred option and it is our main focus in this paper. STAs that transmit successful RTS packets in the TR phase,
all the k STAs transmit their data sequentially according to
II. LITERATURE REVIEW the schedule broadcast from the AP which means there are a
So far, there are some random access schemes for total of k data and k ACK sequential transmissions. Because
OFDMA-based wireless networks that provide random of these k data and k ACKs sequential transmissions for each
access capabilities. The authors in [2] designed a scheme

2166-0662/13 $26.00 © 2013 IEEE 497


DOI 10.1109/ISMS.2013.121
Figure 1. Hybrid-OFDMA (H-OFDMA) Scheme.
Figure 2. Concurrent-OFDMA (C-OFDMA) Scheme.
STA, plus the scheduling overhead, H-OFDMA MAC
protocol results in low throughput in crowded WLAN transmit their data traffic concurrently. The main features of
systems. the C-OFDMA scheme that result in improved throughput
In this paper, we propose a MAC protocol for WLANs are:
based on OFDMA and CSMA/CA to improve the throughput • Concurrent Transmission: Multiple STAs transmit
of the H-OFDMA scheme. We design a Concurrent OFDMA concurrently using multiple STAs and one block ACK.
(C-OFDMA) MAC protocol where STAs transmit their data
This saves the interframe spacing time intervals that
simultaneously after the contention, and save interframe
might have elapsed after each sequential transmission.
spaces. By using concurrent transmission in the ST phase of
H-OFDMA scheme, all the k STAs that have successfully • Efficient scheduling scheme: The AP applies an
transmitted RTS packet send their data at the same time efficient scheduling scheme with little overhead that
using the multiple SCHs and need one block ACK packet. binds the SCHs to the STAs, and this message is
Moreover, we use an efficient scheduling technique that communicated via the block CTS packet.
binds the STAs and the SCHs that they win in the random • Two dimensional back off strategy: The fact that we
access. The AP relays this schedule information in a block have frequency and time resources enable an efficient
Clear-To-Send (CTS) packet, and each STA that contend for back off procedure in C-OFDMA.
an RTS packet learns the success/failure of their contention Fig. 2 shows the proposed C-OFDMA scheme. We will
from the block CTS. describe the detailed working principles of these three phases
The rest of the paper is organized as follows. The of the C-OFDMA below.
proposed MAC protocol will be explained in detail in
Section III. An analytical model for the proposed MAC A. Sub-channel Request (SR) phase
scheme is discussed in Section IV. Performance analysis In SR phase, an OFDMA-based CSMA/CA scheme with
results are presented in Section V to validate the proposed exponential backoff process is implemented for each sub-
protocol. Finally, the paper is concluded in Section VI. channel to handle possible collision. In this phase, we apply
a two dimensional backoff technique. In OFDMA-based
III. PROPOSED SCHEME – CONCURRENT OFDMA CSMA/CA, an STA can obtain the channel information of
We consider a centralized WLAN network that consists all the SCHs by performing fast Fourier transform (FFT)
of one AP at the center and n STAs spread within the range operation on the received OFDMA symbol. This enables
of the AP. We assume that all STAs and the AP are STAs to contend in frequency and time domain. An STA that
synchronized with each other. The total bandwidth of this has packet to transmit senses all the SCHs. If no SCH is idle,
OFDMA-based WLAN network is divided into M narrow- the STA keeps sensing until it finds idle SCH. If at least one
band SCHs. To explain the C-OFDMA model in detail, we SCH is idle for DIFS, a backoff timer is started immediately.
define three phases; Sub-channel Request (SR) phase, Sub- The STA is allowed to transmit only at the beginning of each
channel Assignment (SA) phase, and Data transmission (DT) symbol period. The beginning of the SR phase includes q
phase. During the SR phase, the STAs in the network symbol periods to detect an OFDMA symbol in each of the
randomly contend to send an RTS packet for sub-channel sub-channels. The q symbol periods are used for the
request based on CSMA/CA scheme. Each STA can access exponential backoff process. When an STA wants to
one of the multiple SCHs through contention at each time transmit an RTS packet in the SR phase, it chooses a random
instant. During the SA phase, the STAs that have sent an backoff number from the interval (0, W-1), where W is the
RTS packet in the SR phase learn the result of the random contention window and depends on the number of failed
contention from the broadcast message sent by the AP in a transmissions of the packet. At the first transmission attempt,
block CTS message. After the completion of the SR phase W is set equal to the value of the minimum contention
and SA phases, the DT phase follows where the STAs that window, Wmin. The backoff counter is decremented if each
have received a channel assignment in the SA phase OFDMA symbol is idle and is frozen when an OFDMA

498
TABLE I. SCHEDULE OF THE BLOCK CTS MESSAGE FROM AP
STA ID Result of Block CTS Packet
STA 1 ID SCH 2
STA 3 ID SCH 1
STA 5 ID SCH 3

down to zero at the same time, and they send their RTS
packet in SCH 4 which will result in collision. The STAs
wait for the block CTS packet to know the result of the
contention. The block CTS message that is broadcast from
the AP will contain the information shown in Table I. After
receiving the block CTS broadcast message from the AP,
Figure 3. Example of C-OFDMA Scheme. STA 1, STA 3 and STA 5 learn that they successfully
transmitted their RTS packet on SCH 2, SCH 1and SCH 3
Symbol is detected on the same sub-channel. An STA respectively, and they transmit their data packet in the DT
transmits when the counter reaches zero. phase on these SCHs sequentially. Then, the AP broadcasts a
In Fig. 2, there are n STAs contending for M SCHs in the block ACK message for STA 1, STA 2 and STA 3 on SCH
SR phase. There is no specific correspondence between 2, SCH 1 and SCH 3 respectively. On the other hand, STA 2
STAs and SCHs as the SR is a random phase which means and STA 4 know from the block CTS that their RTS packet
any STA can pick any SCH based on the CSMA/CA scheme failed because of collision. Now, STA 2 and STA 4 update
and its backoff timer. their backoff counter according to the exponential back off
algorithm and retransmit RTS packet in the next frame of the
B. Sub-channel Assignment (SA) phase SR phase. Since collision occurred at SCH 4, it will be idle
After the STAs transmit their RTS packet in the SR until next frame.
phase, the AP decodes the RTS packet to get which STAs IV. ANALYTICAL MODELING
succeeded in transmitting an RTS packet and then relays a
block CTS packet. The block CTS packet is composed of We extend the analytical modeling of the H-OFDMA
the addresses of each of the STAs with successful RTS scheme and modify it to our proposed protocol. The work in
packet and the SCHs they won. Here, we use a simple [10] has modeled the MAC events of successful
scheduling technique to match the SCHs to the STAs. This is transmission, collision, and idle waiting by assuming two
fundamental assumptions. First, after each idle slot, each
a binding between the STAs and the SCHs. Up on reception
station may attempt to transmit with an independent and
of the block CTS message, the STAs check for a sub-channel
constant probability τ. Second, regardless of the number of
assigned to each of them. No response from the block CTS past collisions, a transmission attempt may result in a
packet for any STA is considered as an unsuccessful collision with an independent and constant probability p. The
transmission, and it re-transmits in the next frame by backoff process is then modeled by a two-dimensional
doubling its contention window W up to a maximum value of Markov chain and the following two equations have been
Wmax given by 2 Wmin, where denotes the maximum found for τ and p in [10]:
backoff stage.
C. Data Transmission (DT) phase , (1)
In this phase, the STAs that have got an SCH in the SA
message from the block CTS packet transmit their data. 1 1 , (2)
SCH/SCHs that resulted in collision in the SR phase will be where n is the number of contending stations, W is the
idle during the DT phase until the next subframe. We minimum contention window, and m is defined so that
assumed that each of the STA have a single data packet to
send in the DT phase carried by one SCH. After the data Wmax = W2m. By solving these equations using numerical
packet is transmitted, the AP uses that SCH to send the methods, one can find τ and p based on the known values
ACK. In this way, a block ACK is sent to all the STAs that of W, m, and n. Using τ and , we calculate the probability
transmitted data. that at least one transmission happens in a given slot (Ptr)
Consider a model where there are 5 STAs and 4 SCHs in and the probability of a transmission in a slot being
the WLAN system as shown in Fig. 3. After sensing the
successful (Ps). Ptr is the complement of the probability of
whole sub-channels idle for DIFS interval, these 5 STAs
randomly contend for the 4 SCHs in the SR phase. Let’s no station transmitting ( 1 ) and is given by:
assume STA 1 sends its RTS packet in SCH 2, STA 3 in
SCH 1, and STA 5 in SCH 3 based on their backoff counters. 1 1 . (3)
There were no other STAs that reached their backoff timer The probability of a transmission in a slot being
zero at the same time with STA 1, STA 3 and STA 5. On the successful, Ps, is the probability of exactly one transmission
other hand, the backoff timers of STA 2 and STA 4 counted given that there has been a transmission on the channel, and

499
is given by: where E[PT] is the Expected duration for each STA in the
P . (4) parallel transmission, and E[IFS] is the expected duration of
the interframe spaces of the DT phase. E[PT] is given by:
The saturated throughput analytical modeling for the C-
OFDMA scheme is developed based on these values. In the 2 , (10)
C-OFDMA scheme, the above probabilities describe the where and ℎ are the length of the MAC and
probabilities that each of the q backoff slots contain a physical headers, and and are the physical operational
transition to successful collision or idle states, and the value and basic rates.
of n changes to the number of STAs assigned to each SCH. The value of E[IFS] is calculated as follows:
The probability that an SR phase sub-channel is successful in
delivering a request, denoted as P SR , can be found μ 2 δ , H OFDMA
considering that a successful transmission can happen at any E IFS ,
2 2 δ, C OFDMA
slot during an SR phase assuming that its previous slots were (11)
i−1 th
idle ((Pidle) for i slot). This can be written by summing where δ is the propagation delay, denotes the time
th
the probabilities of the i slot being successful and all (i-1) duration of the DIFS, and denotes the time durations of
previous slots empty. the SIFS, In this alone, there is a save of μ
μ 2 in C-OFDMA compared to H-OFDMA.
∑ . (5) The expected length of the SR phase, E[SR] is
approximated as the duration of one RTS message plus half
Similarly, the probability of an SR phase sub-channel being of the backoff slots, plus the Expected duration of
idle is given as the probability that all q slots are idle: transmitting the CTS message, E[CTS], of length LCTS
(TCTS). The random backoff number is uniformly chosen
. (6) from the contention window, and the length of the SR
message sent using OFDMA is stretched due the use of M
Now, the expected throughput of the C-OFDMA is sub-channels, same us that of H-OFDMA. Thus,
calculated by dividing the expected amount of traffic
delivered in each DT phase, by the expected duration of the E SR δ, (12)
DT phase (E[DT]) plus the expected duration of the SR where LSR is the length of an SR message in bits and is equal
phase (E[SR]) plus the expected duration of the SA phase to ( ) and Tslot denotes the duration of backoff slots in the
(E[SA]). We assume that all STAs contain a single packet beginning of each SR phase.
transmission in the DT phase. The expected amount of data The expected value of the SA phase, E[SA], is the
served in the DT phase is found by multiplying the expected expected number of transmitting the CTS message of length
length of a packet (E[P]) by the expected number of LCTS (TCTS), and is given by:
successful SR sub-channels (µ). The value of µ is computed
by multiplying the number of sub-channels used in the / μ / δ. (13)
system by the probability of successful SR phase sub-
channel, . Based on the number of STAs in the network, The achievable throughput of H-OFDMA is derived in
there are two cases. Case 1 is when the number of STAs [9]:
assigned to all SCHs is equal, and Case 2 is when the
, (14)
number of STAs assigned to all the SCHs is not equal. Based
on these two cases, µ is computed as follows: where E[TR] and E[ST] are given by:

MP SR , Case 1
μ , (7) / / /2
∑M P , Case 2
where is the of the ith sub-channel. With all the δ. (15)
above values at hand, the throughput of the proposed C-
OFDMA MAC protocol is given by: μ
δ
EP
SC OFDMA . (8) μ 2
E SR E SA E DT

The Expected length of the DT phase is given by: μ 2δ . (16)

μ , (9) Finally, the throughput, S, of IEEE 802.11 DCF and


RTS/CTS are derived in [10], and given as follows:

500
S . , (17)

where Ts is the average time the channel is sensed busy


because of a successful transmission, and Tc is the average
time the channel is sensed busy by each station during
collision busy by non-colliding stations. For 802.11 DCF, the
values of Ts and Tc are given as follows:
= 2 / ( )/
2δ ,
= / ( + )/ + + δ,

For 802.11 with RTS/CTS, they are given as:


=4 / ( )/R
3 4δ ,
= / ( )/ δ,
where TDIFS time durations of the RIFS, and all the other Figure 4. Saturated throughput comparison of DCF, RTS/CTS
and C-OFDMA MAC schemes, M = 16
parameters are same as defined in the earlier equations. Now,
we can compare the throughputs of C-OFDMA with DCF, because there are more number of sub-channels than there
RTS/CTS, and H-OFDMA. are STAs, which causes resource underutilization, and hence
throughput loss. This can be solved by adaptively changing
V. SIMULATION RESULTS
the number of sub-channels if the number of STAs is smaller
In this section, we conduct performance analysis based than the number of sub-channels in the system. As the
on the analytical modeling to show the validity of the number of STAs increases, the throughput of the C-OFDMA
proposed MAC protocol. Table II shows the parameters keeps increasing while the other schemes decrease as shown
used in the simulation. in Fig. 4. The increase is due to the higher data rate and
Fig. 4 depicts the comparison of the saturation lower overhead of the concurrent data traffic, block CTs and
throughput of the basic IEEE 802.11 DCF, and extended ACK packets.
IEEE 802.11 RTS/CTS single channel protocols with the Fig. 5 shows the saturated throughput of the H-OFDMA
multichannel C- OFDMA scheme based on the parameters and C-OFDMA MAC protocols when M is 16 and 32. In
given in Table II. We observe that the multichannel C- both cases, C-OFDMA provides higher throughput than H-
OFDMA obtains a higher throughput than the single channel OFDMA because of the simultaneous data, and block ACK
DCF and RTS/CTS MAC protocols because multiple STAs transmissions. For example, when the number of STAs in
can transmit simultaneously by using the multiple sub- the network is 60 (n = 60), C-OFDMA protocol provides
channels. For example, when the number of STAs in the 16% improvement over H-OFDMA. For the both H-
network is 60, C-OFDMA results in a 74 % and 45 % OFDMA and C-OFDMA protocols, M = 32 provides smaller
throughput improvement over the IEEE 802.11 DCF, and throughput when the number of STAs in the network is
IEEE 802.11 RTS/CTS single channel schemes respectively. smaller than the number of sub-channels (n < M). This is
When the number of STAs is small, the throughput of the
C-OFDMA scheme is lower than the other schemes. This is
TABLE II. PARAMETERS AND THEIR VALUES USED
Parameter Value Parameter Value
BW 20 MHz ACK Length 96 bits
FFT Size 256 CTS Length 144 bits
Effective Symbol 12.8 micor RTS Length 144 bits
Time sec
Guard Time 3.2 micro sec Hmac 240 bits
Slot Time 16 micro sec Hphy 120 bits
SIFS 10 micro sec M 16, 32
DIFS 20 micro sec m 4
RIFS 30 micro sec CWmin 16
Propagation 1 micro sec Data 8184 bits
Delay, δ Length
q 8
Figure 5. Saturated throughput comparison of H-OFDMA, and C-
OFDMA for M = 16 and M = 32.

501
then transmit their data packets concurrently by using their
corresponding SCHs. Performance analysis shows that C-
OFDMA saves significant interframe spaces and results in
improved throughput in crowded WLAN environment.
From physical layer point of view, there might be
interference among the different SCHs during the DT phase
which needs a guard band. In this paper, we assume that the
AP can dynamically use some of the SCHs that are idle due
to collision for this purpose. Another interesting point is to
devise a mechanism that can wisely use the SCHs that
remain idle in DT phase because of the collision they
encountered in the SR phase for data transmission. We will
consider these issues in our future work.
ACKNOWLEDGEMENT
This research was supported by the MKE (The Ministry of Knowledge
Economy), Korea, under the ITRC (Information Technology Research
Figure 6. Saturated throughput comparison of H-OFDMA, and Center) support program supervised by NIPA (National IT Industry
C-OFDMA for SIFS = 10 and M = 16. promotion Agency (NIPA-2012-(H0301-12-2003)).

because of channel underutilization that causes some sub- REFERENCES


channels to be idle. When n > M (STAS greater than number [1] IEEE std. 802.16e-2005, IEEE Standard for local and metropolitan
of sub-channels), the throughput increases because channel area networks- part 16: Air Interface for fixed broadband wireless
utilization improves. Another reason is because the number access systems amendment for physical and medium access control
of successful sub-channels increases and the concurrent layers for combined fixed and mobile operation in licensed bands,
transmission in the DT phase saves the interframe spacing Dec. 2005.
time intervals. [2] X.Wang and W. Xiang, “An OFDM-TDMA/SA MAC protocol with
In Fig. 4 and Fig. 5, we used the parameters given in QoS constraints for broadband wireless LANs,” Wireless Networks,
vol. 12, no. 2, pp. 159–170, 2006.
Table II to compare the performance of the proposed scheme
[3] S. Valentin, T. Freitag and H. Karl, ”Integrating Multiuser Dynamic
with H-OFDMA. As can be seen in Table II, the interframe OFDMA into IEEE 802.11 WLANs - LLC/MAC Extensions and
spacing values are SIFS = 10 micro sec, DIFS = 20 micro System Performance,” in Proc. IEEE Global Interational Conf.
sec, and RIFS = 30 micro sec. But, we can use different Commu. ICC, May 2008.
interframe space values larger than that. In Fig. 6, we use, [4] H-S. Ferdous and M. Murshed, ”Enhanced IEEE 802.11 by
SIFS = 16 micro sec which is common in wireless networks. integrating multiuser dynamic OFDMA,” in Wireless Telecomm.
In this case, C-OFDMA provides significant improvement in Symposium WTS, Apr. 2010.
throughput when compared to H-OFDMA because when the [5] Y. Choi, S. Park, and S. Bahk, “Multichannel random access in
system uses larger interframe space values, the proposed OFDMAwireless networks,” IEEE J. Sel. Areas Commun., vol. 24,
no. 3, pp. 603–613, 2006.
concurrent transmission scheme becomes very efficient. For
example, when number of STAs n = 60, and SIFS = 16, C- [6] H. Kwon, H. Seo, S. Kim and B-G. Lee, ”Generalized CSMA/CA
Protocol for OFDMA Systems,” in Proc. IEEE Global Telecommun.
OFDMA provides 24% throughput improvement over H- Conf. GLOBECOM, Nov. 2008.
OFDMA scheme in Fig. 6. We can observe that when we [7] H. Kwon, H. Seo, S. Kim, and B.-G. Lee, “Generalized CSMA/CA
apply C-OFDMA protocol, if the interframe spaces are large, for OFDMA systems: protocol design, throughput analysis, and
the throughput improvement also becomes proportionally implementation issues,” IEEE Trans. Wireless Commun., vol. 8, no.
large. 8, pp. 4176–4187, Aug. 2009.
[8] H. M. Alnuweiri, Y. P. Fallah, P. Nasiopoulos, and S. Khan,
VI. CONCLUSION “OFDMA-based medium access control for next-generation
WLANs,” URASIP J.Wireless Commun. and Networking, vol. 2009,
In this paper, C-OFDMA, a new MAC protocol for article ID 512865, 9 pages, 2009.
WLAN is proposed based on OFMDA and CSMA/CA. The [9] Y. P. Fallah, S. Khan, P. Nasiopoulos, and H. M. Alnuweiri, “Hybrid
main feature of the C-OFDMA protocol is that the presence OFDMA/CSMA based medium access control for next-generation
of multiple sub-channels in OFDMA allows many STAs to wireless LANs,” in Proc. 2008 IEEE Int’l Conference on Commun.,
transmit their data simultaneously after they receive pp. 2762–2768.
successful RTS packet in the random access phase. Using an [10] G. Bianchi, “Performance analysis of the IEEE 802.11 distributed
efficient scheduling technique, the AP broadcasts a block coordination function,” IEEE Journal on Selected Areas in
CTS packet that binds the STAs and the SCHs. The STAs Communications, vol. 18, no. 3, pp. 535–547, 2000

502

You might also like