master’s thesis resource allocation scheme for spectral...
TRANSCRIPT
Master’s Thesis
Resource Allocation Scheme for Spectral Efficiency Enhancement of OFDMA
WLAN Systems
Department of Electrical and
Computer Engineering
Graduate School of Ajou University
December, 2016
Moonmoon Mohanty
Resource Allocation Scheme for Spectral Efficiency Enhancement of OFDMA
WLAN Systems
Supervisor: Professor Jae-Hyun Kim
by
Moonmoon Mohanty
A thesis submitted to the Graduate School of Ajou University
in Partial Fulfillment of the Requirements for the Degree of
Master of Science
Department of Electrical and
Computer Engineering
Graduate School of Ajou University
December, 2016
Master’s Thesis of Moonmoon Mohanty is hereby approved by the guidance
committee:
Graduate School of Ajou University
December 20th, 2016
i
Acknowledgement
Firstly, I would like to express my genuine gratitude towards my
supervisor, Professor Jae-Hyun Kim, for his support, guidance and constant
encouragement. Professor Jae-Hyun Kim cared about me a lot, not only in
academic status but also about living in Korea. Professor Jae-Hyun Kim also
gave me a chance to participate in a project from which I experienced
significant advantages. His entertainments were also additional and provided
great chances to enjoy living in Korea and to understand its culture. Once
again I am very thankful to Professor Jae-Hyun Kim for all what he did for me
and for his unlimited kindness.
Besides my supervisor, I would also like to thank the other committee
members: Professor Kyo-Beum Lee and Professor Songnam Hong for their
encouragement and valuable comments.
I would like to extend my sincere thanks to the devoted and diligent
members and alumni of WINNER Lab. Thank you so much for answering all
my queries, giving me courage and respect. They all are my good friends and
I spent a great time in WINNER Lab with them. Dr. Kwang-Chun Go, Sung-
Hyung Lee, Hye-Rim Cheon, So-Yi Jung, Jin-Ki Kim, Nathnael
Gebregziabher Weldegiorgis, Hyun-Ki Jung, Jong-Mu Kim, Won-Kyung
Kim, Dong-Yeol Choi, Seung-Su Yoo.
I would also like to thank Chen Yiliang from Communication System
Lab, who has been a great friend and support.
In addition, I would like to sincerely thank my friends and family.
Without their friendship and constant words of support and encouragement,
my two year course would have been extremely hard.
ii
Abstract
As the number of users of wireless devices, not only smartphones but also
tablets are increasing exponentially, WLANs will be playing a crucial role in
next generation wireless communications. The MAC protocol and MA scheme
play a very important role in determining the efficiency and scalability of a
WLAN. In comparison to other access technologies such as TDMA and
CDMA, OFDMA is emerging as a promising solution for wireless networks.
OFDMA assigns sets of subcarriers to different users, this is how multiple
access is achieved. As a result, simultaneous data transmission can take place
from several users. Moreover, OFDMA can overcome one of the major
drawbacks in wireless networks, frequency selective fading.
The existing MAC schemes for wireless local area networks lack proper
utilization of idle or unused sub-channels. One of the key challenges is an
efficient radio resource management which can exploit the channel bandwidth
to the maximum extent. In this thesis we propose resource allocation scheme
based on grouping and fragmentation for improvement of spectral efficiency
in OFDMA WLAN environment. The scheme is applicable to systems with
stations of heterogeneous packet lengths. Earlier the studies have been done
in systems with homogeneous packet lengths. However, packet lengths are
quite heterogeneous in real networks. We demonstrate the effectiveness of the
schemes in handling heterogeneous packet lengths and improving spectral
efficiency and compare it with existing ones through simulation.
iii
Table of Contents
Acknowledgement………………………………………………………….. i
Abstract………………………………………………………………........... ii
Table of Contents……………………………………………………............ iii
List of Figures…...………………………………………………………….. v
List of Tables……….……………………………………………………….. vii
Chapter 1.Introduction…………………………………………………….. 1
1.1 Background and Motivation................................................ 1
1.2 Contributions........................................................................ 3
1.3 Overview….......................................................................... 3
Chapter 2.OFDMA based MAC protocols………………………………... 5
2.1 Overview of OFDMA……………………............................. 5
2.2 Hybrid OFDMA/CSMA Based MAC...................................... 7
2.2.1 Hybrid MAC design………………………………….. 8
2.3 Novel DCF-based MAC........................................................ 10
2.3.1 Novel-DCF protocol.................................................... 10
2.3.2 Performance evaluation……………………………... 14
2.4 C-OFDMA…........................................................................ 14
2.5 OMAX protocol..................................................................... 16
2.5.1 Protocol design……………………………………… 17
2.5.2 Performance evaluation……………………………... 19
2.6 OFDMA based Multiple access protocol…………………... 19
Chapter 3. Resource Allocation Scheme………………………………… 22
3.1 System Model………………………………………………. 22
3.2 MAC protocol procedure……………………………............ 23
3.3 Grouping based resource allocation…………………............ 24
iv
3.3.1 Grouping based resource allocation algorithm........... 25
3.3.2 Example of proposed scheme.................................... 26
3.4 Fragmentation based resource allocation................................ 27
3.4.1 Fragmentation based resource allocation algorithm.... 27
3.4.2 Example of proposed scheme……………………….. 31
3.5 Frame Structure………..…………………………………..... 32
3.6 Fragmentation………………………….…………................. 33 Chapter 4. Performance Evaluation…………………………………….. 35
Chapter 5. Conclusion…………………………………………………… 45
References………………………………………………………….. 46
v
List of Figures
Figure 1. (a) OFDM (b) OFDMA.………......................................... 6
Figure2. Sub-Channelization (a) ASM Method (b) DSM Method… 7
Figure 3. Hybrid MAC operation ………….…………………….... 9
Figure 4. Novel-DCF resource allocation procedure …………....... 11
Figure 5. Optimum Fragmentation Size Determination………...….. 13
Figure 6. C-OFDMA Scheme.…………………………………….. 15
Figure 7. OMAX protocol procedure.………………………...…… 17
Figure 8. QoS-OFDMA procedure……………………………........ 20
Figure 9. Basic Service Set for infrastructure mode…….…............ 22
Figure 10. Grouping based MAC protocol procedure…...…............ 24
Figure 11. Example of grouping based resource allocation…………. 25
Figure 12. Flowchart of proposed resource allocation algorithm…... 28 Figure 13. Data transmission (a) without fragmentation (b) with fragmentation……………………………………………………….. 31
Figure 14. Modified Frame format…………………………………. 33
Figure 15. Fragmentation of MSDU……………………………….. 34
Figure 16. Throughput vs increasing number of stations…………… 39 Figure 17. Throughput vs PHY data rates and number of stations (grouping)…………………………………………………………… 40
vi
Figure 18. Throughput vs PHY data rates and number of stations (fragmentation)…………………………………………....................
41
Figure 19. Throughput vs number of sub-channels ………………… 42 Figure 20. Simulation vs Analytical results (grouping)………….…. 43 Figure 21. Simulation vs Analytical results (fragmentation)……….. 44
vii
List of Tables
Table 1: Modulation and coding scheme in 20MHz …….…………...……35
Table 2: Simulation parameters……………………………………………36
1
Chapter 1
Introduction 1.1 Background and Motivation
As the number of users of wireless devices is increasing exponentially, not
only smart phones but also tablets are increasing exponentially, WLANs
(wireless local area networks) will be playing a crucial role in next generation
wireless communications. There is growing demand for higher data rates and
quality of QoS(quality of service) requirement. In past years, there has been
much interest in the design of WLANs [1], [2]. IEEE 802.11 was formed to
bring an international standard for WLANs. It is a set of MAC(media access
control)and PHY(physical) specifications for implementation of WLAN
communication. DCF(distributed coordination function) is the primary MAC
technique of 802.11 [3]. However, it was seen that with increase in number of
STAs(stations), the throughput decreases. The MAC protocol and
MA(multiple access) scheme play a very important role in determining the
efficiency and scalability of a WLAN. In comparison to other access
technologies such as TDMA(time division multiple access) and CDMA(code
division multiple access) [4-6], OFDMA(orthogonal frequency division
multiple access) is emerging as a promising solution for wireless networks. In
TDMA, since the bandwidth is shared in time, strict synchronization is
required. Therefore, when CSMA(carrier sense multiple access) is used with
TDMA, efficiency of the system will be reduced. In CDMA, high bit-rate
pseudo-code sequence is multiplexed with the signal to be transmitted. The
2
drawback of this transmission is its susceptibility to channel frequency
selectivity. Moreover, in order to recover from ISI(inter symbol interference),
very complex channel equalization is required at receiver end. However,
OFDMA is based on OFDM(orthogonal frequency division multiplexing).
OFDM has resilience to fading effects of the wireless channel. ISI and
expensive channel equalization can be completely avoided in OFDM systems.
This makes OFDM more suitable for WLANs. In OFDMA, subcarriers are
grouped into SCHs(sub-channels) and assigned to individual users. The other
advantages of OFDMA are scalability and flexibility of deployment.
One of the main characteristics of the next generation WLAN is dense
deployment. However, the performance of existing MAC protocols degrade in
crowded networks and are not efficient to support heterogeneous traffic types.
This is due to single user channel access and single user data transmission. But
with advent of OFDMA, more efficient MAC protocols can be devised, since
it enables multiuser channel access and multiuser transmission. Some recent
works on OFDMA WLANs have been presented in [7-11]. In [7-9], the
authors presented a MAC protocol that combines OFDMA with
CSMA/CA(CSMA with collision avoidance). They investigated the effect of
increasing load on performance and found notable improvement in throughput
compared to pure CSMA system. The authors in [10], have presented a C-
OFDMA(concurrent OFDMA) MAC protocol and compared it with existing
protocols. Due to the presence of multiple SCHs, STAs can transmit data
concurrently once they receive RTS(request to send) packet. This results in
better performance. However, there is no mechanism to utilize the idle SCH
that exist due to collision. In [15], [16], there has been a survey on the existing
OFDMA MAC protocols.
3
There has been extensive research on channel access schemes for OFDMA
WLANS in [11-14]. Most of them focus on throughput improvement due to
multi-user diversity gain. Some other works [17], [18] concentrate on QoS
satisfaction. Many protocols have been proposed in [21-28]. However, none
of the current researches found a solution to existence of idle channel. Due to
idle channels, the spectral efficiency of the system degrades.
1.2 Contributions This thesis proposes a resource allocation algorithms based on grouping and
fragmentation, to improve the spectral efficiency of the system by reducing
radio resource wastage. The system considered has stations with
heterogeneous packet lengths.
The key objective of our protocol is to address the issue of unused channel.
In our system, the AP(access point) acts as central controller and uses the
traditional DCF method to block radio resource. The groups are formed as per
the algorithm. When the number of stations in Gth group is less than the
number of groups, the AP implements the proposed resource allocation
algorithm and allocates the unused segment of sub-channels to the fragments
of the packet of required station. This results in better spectral efficiency.
1.3 Overview
The remaining part of the thesis is organized as follows. In Chapter 2, an
overview of OFDMA system and related works is described. We propose an
overall resource allocation algorithm by implementing grouping and
fragmentation to solve the problem of idle channel or resource wastage in
4
Chapter 3. Performance evaluation of above algorithms is represented in
Chapter 4. The thesis is concluded in Chapter 5.
5
Chapter 2
OFDMA Based MAC protocol 2.1 Overview of OFDMA OFDMA is a system where resources can be assigned in both time and
frequency domain. Frequency selective fading is a huge challenge in WLANs.
It is caused by multi-path propagation. OFDMA can address this problem in
an elegant manner. There are many advantages of OFDMA which make it
suitable for wireless networks – scalability, ability to take advantage of
channel frequency selectivity and use of multiple antennas MIMO(multiple
input multiple output) -friendliness. There are other advantages of OFDMA as
mentioned in [20]–
Deployment flexibility across various frequency bands.
Multi-user diversity
The process is simplified at the receiver end, FFT(fast fourier
transform) processor is required.
Spreads the carriers all over the used spectrum, thus providing
frequency diversity.
Figure 1 shows how signals are allocated time and frequency resources. In
order to separate transmissions from multiple users, groups of OFDM
symbols/subcarriers are used. Hence, the subcarriers are the smallest
allocation units in frequency domain and OFDM symbol period in time
6
domain. In OFDM, only one user is allocated all the sub-carriers at a time.
However, in OFDMA different users can be allocated different sub-carriers at
any given time.
(a) (b)
Figure 1: (a) OFDM vs (b) OFDMA
Grouping of subcarriers in an OFDM symbol in different ways leads to
creation of SCHs. There are types of sub-channelization as seen in Figure 2 –
1. ASM(adjacent subcarrier method):
A contiguous group of subcarriers is mapped into a sub-channel.
2. DSM(diversity subcarrier method):
Grouping is based on permutation or diversity. The sub-channel
contains non-contiguous subcarriers.
7
IFFT(inverse fast fourier transform) is implemented at the transmitter end
and FFT at the receiver end. By performing FFT on the received OFDMA
symbol, a STA can collect information of all the SCHs in an OFDMA system.
Figure 2: Sub-Channelization Examples: (a) ASM method, (b) DSM method.
2.2 Hybrid OFDMA/CSMA Based MAC
The motivation behind this work was low MAC layer efficiency in crowded
networks. The authors have presented a MAC protocol which implements
OFDMA with CSMA/CA scheme. The frame delivery takes place in two
stages [7] – 1. TR(transmission opportunity request) phase, 2. ST(scheduled
data transmission) phase. IFS(inter frame space) is the time that separates the
two phases. Here there are three IFS values. MIFS(minimum inter frame space)
8
is the time required by the PHY and the MAC to receive and process the
frame’s last symbol. The IFSs are determined by CCA(clear channel
assessment) mechanism. CIFS(controlled access inter frame space) = MIFS +
CCA, RIFS(random access inter frame space) = CIFS + CCA. Hence, MIFS
< CIFS < RIFS.
2.2.1 Hybrid MAC Design
The AP sends a TR start message, following which the TR phase begins after
MIFS interval. The TR phase uses OFDMA, where each station is assigned a
sub-channel to send TR message. The TR phase will be contention free if the
number of stations is equal to or less than the number of sub-channels. In order
to resolve collisions, CSMA/CA is used. Thus in TR phase hybrid
OFDMA/CSMA is used. The transition from TR phase to ST phase happens
after the AP has scheduled the ST phase. First, the channel has to be sensed
idle for CIFS time. Then the stations can transmit data as per the schedule
broadcasted by the AP. However OFDMA is not utilized in the ST phase.
Figure 3 [7], depicts the entire operation. The two phases are described in
details below –
1. TR phase:
Each station has to contend for sub-channel, to transmit TR message.
CSMA/CA is used for contention in every sub-channel. A random
exponential backoff process is used for resolving collisions. Any
station willing to send a message in the TR phase, will choose a
9
backoff number. It is selected randomly from the interval (0, ),
where CW(contention window). When the slot is empty, the backoff
counter will be decremented. Once the backoff counter reduces to zero,
the TR message is sent. Then the station will wait for AP’s reply. In
case there is no reply, the station considers it as a collision.
Figure 3: Hybrid MAC operation.
This is followed by doubling of contention window and selection of a
new random backoff number.
2. ST phase:
AP sends the first message in this phase. It notifies the order of access
and transmission schedule to all the stations by broadcasting the ST
schedule. This is followed by CIFS spacing and then data transmission
by stations in the specified order. Following the data frame, after MIFS
spacing, ACKs(acknowledgements) are sent.
10
For performance evaluation of the Hybrid MAC, normalized throughput was
computed and compared with pure CSMA. The packet length was assumed to
be 2000 bytes. From simulation results, it was found that the Hybrid MAC
achieved almost 30% gain in performance.
2.3 Novel DCF-Based MAC
The motivation behind this protocol is throughput enhancement. The BSS
considered here operates in infrastructure mode. In this protocol, the radio-
resource is reserved by the AP, for a fixed duration according to conventional
DCF. Then the AP collects required information from stations and allocates
RBs (resource blocks) through two-dimensional scheduling. The resource
allocation procedure constitutes the following steps:
Step 1. Generation of RBs.
Step 2. Sub-channel allocation to STAs.
Step 3. Fragmentation of DATA Frame.
Step 4. Allocation of RBs.
Step 5. Update of Resource Reservation Duration.
2.3.1 Novel-DCF Protocol
Figure 4[11], depicts the steps followed in resource allocation of Novel-
DCF. This protocol is applied to WLANs consisting of two types of traffic –
RT(real time) and NRT(non real time) traffic. Once the radio resource is
reserved, the remaining resource reservation period is found which is used for
11
actual data transmission. It is the total resource reservation period – the control
frame transmission period.
Figure 4: Novel-DCF resource allocation procedure.
Step 1. Generation of RBs:
The AP calculates the number of resource blocks in time domain/number
of time slots, by
12
(1)
where is the duration of 1 RB, is the remaining resource
reservation period. The number of resource blocks is the product of
number of SCHs and the number of time slots found in (1).
Step 2. SCH allocation to STAs:
The AP selects RT station on the basis of EDF(earlier deadline first)
scheduling scheme. The RT station with smallest TTL(time to live) value
is allocated one OFDMA SCH with highest SNR(signal-to-noise ratio) to
the RT station. After the RT STAs are allocated SCHs, the remaining
unallocated SCHs are assigned to NRT stations.
Step 3. Fragmentation of DATA Frame:
After sub-channel allocation, there may be some unallocated RBs as
every station has different amount of data to transmit. As seen in Figure
4(d), there are many unallocated RBs in SCH#1, 3 and 4. In order to solve
this problem, fragmentation has been used. An optimum fragmentation
size has to be determined. The fragmentation size should be such that the
number of unallocated RBs in the upcoming data frame is maximized.
The technique of finding the optimum fragmentation size is depicted in
Figure 5. As seen in the figure, when fragmentation size = 4, the new
data frame has maximum number of RBs unallocated to any STA. Hence
that is the optimum fragmentation size.
Step 4. Allocation of RBs:
13
After fragmentation, RBs are allocated to new data frames.
Figure 5: Optimum Fragmentation Size Determination.
Step 5. Resource Reservation Duration Update:
is updated after allocation of RBs is complete. The new
resource reservation period is given by
14
= – ( + + ) (2)
where is the duration of preamble, is the SIFS(short inter
frame space) duration and is the duration of preceding frame.
2.3.2 Performance Evaluation
For the simulation, the BSS(basic service set) considered consists of one
RT STA and N NRT STAs. The RT STA generates packets of 200byte and
NRT STA generates 1500 byte packets. Simulation results confirm that the
protocol enhances throughput of WLAN.
Problems:
Since the radio resource is reserved for a fixed duration, with increasing
number of stations, number of control frames increase. As a result the time for
which actual data transmission is allowed, decreases. The resource allocation
method is complex due to fragmentation process.
2.4 C-OFDMA
This protocol was proposed for throughput improvement of Hybrid
OFDMA [7]. As per this protocol, during contention phase all STAs that send
RTS successfully, can transmit data concurrently to the AP. The WLAN
considered here, contains one AP and n STAs. It is assumed that all the STAs
15
and AP are synchronized. Figure 6 [10], depicts the entire operation. The three
phases are described in details below –
Figure 6: C-OFDMA Scheme.
1. SR(sub-channel request) phase:
Each sub-channel implements its own CSMA/CA scheme. The STA
having packet for transmission first senses all the SCHs. It keeps
sensing until it comes through any SCH that is idle. A backoff timer is
started in case any SCH is idle for a duration of DIFS.
The STA selects a random backoff number from (0, W-1). W stands
for contention window. The STA can transmit RTS once the backoff
counter reaches zero. As shown in Figure 7, n STAs are contending for
M SCHs.
2. SA(sub-channel assignment) phase:
16
The AP processes the RTS packets to find the STAs successful in
RTS transmission. The AP sends a Block CTS(clear to send). It is a
map containing STA ID and its corresponding SCH. When the STAs
receive this CTS, they look for their allocated SCH. No information
regarding SCH implies unsuccessful transmission, and the STA will
have to retransmit in next round by doubling its contention window.
3. DT(data transmission) phase:
The STAs transmit their data packet on the SCH assigned to them in
SA phase. The AP sends ACK on the same SCH. There is a block ACK
sent to all STAs.
The saturation throughput of C-OFDMA was compared with IEEE 802.11
RTS/CTS and Hybrid OFDMA [10]. It was seen that C-OFDMA had higher
throughput compared to others. This is because multiple STAs can transmit
simultaneously on multiple SCHs.
Problems:
Idle SCHs due to collision in SR phase.
2.5 OMAX protocol(OFDMA based multiple access for
802.11ax)
WLANs have been playing an important role in our lives, providing high
speed connectivity. The peak physical rate has been increasing over the years.
In March 2013, a new WLAN standard was set up, IEEE 802.11ax. This
17
standard is expected to improve average throughput by four times. This can be
achieved by implementing OFDMA. The major challenges that need to be
overcome for implementing OFDMA are - overhead reduction and
synchronization. The authors have suggested fast backoff and whole channel
physical sensing, in order to overcome synchronization problem. They have
also proposed an enhanced RTS/CTS mechanism and frame structures to
reduce overhead.
2.5.1 Protocol Design
Figure 7: OMAX protocol procedure.
The procedure of OMAX is illustrated in Figure 7 [17].
1. Whole channel physical channel sensing:
18
Earlier, STAs used to sense each SCH to check if it is idle [10].
However in OMAX, all STAs check the entire channel until it is idle
for DIFS(DCF inter frame space). The state is considered idle if all the
SCHs are idle, else it is considered busy. Therefore, if one STA
transmits RTS, the other STAs consider the channel state as busy. If
the backoff counter of STAs reach zero simultaneously, then they can
start transmission simultaneously.
2. Fast Backoff Process:
Here, each STA has only one backoff counter for all SCHs.
Whenever there is an idle slot, the backoff counter is reduced by N,
the number of SCHs. The STA can transmit only when the backoff
counter reaches zero.
3. Enhanced RTS/CTS method:
Usually when STAs send RTS simultaneously, then AP doesn’t
receive any of them due to collision. But here when multiple STAs
transmit their RTS concurrently, the AP receives some of them. After
receiving RTS packets, AP implements scheduling algorithm for
subcarrier assignment. AP informs the STAs about the schedule by
transmitting a G-CTS(group CTS). After receiving G-CTS, the STAs
can transmit data simultaneously, only on their allocated SCHs. The
AP sends acknowledgement for all STAs in G-ACK(group ACK).
4. Frame structure changes:
19
Some changes have been made to the frame structures of G-CTS and
G-ACK. The G-CTS, consists of an additional SI(scheduling
information) field, which contains the SCH information. There are
more than one RA(receiver address) fields as well. In the G-ACK
frame, a new field ACK info has been introduced. The ACK info field
is used to acknowledge all the packets received by AP.
2.5.2 Performance Evaluation
The performance of OMAX has been compared to that of DCF. It was seen
that throughput of OMAX was better than that of DCF.
2.6 OFDMA based Multiple Access Protocol with QoS
Guarantee
This protocol aims to guarantee video traffic QoS. Video traffic requires low
delay and delay jitter. As soon as it is generated, it needs to be transmitted to
the AP in a small duration. The BSS considered here consists of an AP and
stations. There are two types of stations – video and background. The
transmission process is illustrated in Figure 8 [18]. It consists of 3 phases –
1. Contention phase:
The background stations use the same contention method as OMAX
[17]. But the technique is different for video stations. The video station
20
transmits n Q-RTS(QoS-RTS) packets on n arbitrary SCHs when the
backoff counter value is less than N.
2. Resource Allocation phase:
The resource allocation is performed by the AP on the basis of traffic
types. When AP receives Q-RTS packets, it checks the traffic type. If
at least one of the STAs is video type, the AP assigns the entire channel
Figure 8: QoS-OFDMA procedure.
to that STA. Otherwise it divides the whole channel into k SCHs,
where k is the number of successful Q-RTSs received by the AP.
3. Transmission phase:
21
The STAs transmit data on their allocated SCHs as mentioned in G-
CTS and AP replies with G-ACK.
Simulation results show that this protocol performs better than OMAX
protocol. Even the delay and delay jitter of video traffic is lower than that of
OMAX.
22
Chapter 3
Resource Allocation Schemes
3.1 System Model
In the following explanation, we consider a BSS consisting of 1 AP and
N stations (STA1 to STA N) and four OFDMA sub-channels (SCH#1 to
SCH#4). However, our proposed algorithm can also be applied for other
values of K, number of SCHs. We have made certain assumptions for our
environment (i) Heterogeneous packet length i.e. each station has a packet
length different than others. (ii) Saturation condition i.e. stations always have
data available in their transmission queue. (iii) Homogeneous data rate i.e.
stations employ same data rate. The BSS used is shown in Figure 9.
Figure 9: Basic Service Set for infrastructure mode.
23
3.2 MAC Protocol Procedure
There are three phases as seen in Figure 10 [19]–
1. RTS/CTS exchange phase:
The transmission begins with the exchange of RTS/CTS frames. First,
the AP broadcasts RTS frame, which contains the STAs’ order of reply.
All the N stations send their respective CTS in the same
2. DL/UL Resource Allocation phase:
Following this exchange, the AP executes the resource allocation
algorithm and transmits the DL-RAI(down-link resource allocation
information) to all STAs, in order to notify assigned sub-channel and
transmission time. The next step is data transmission by AP on the
allocated sub-channels. Then, the stations inform the AP about their
respective packet size as well as acknowledgements, by transmitting
DL-ACK(downlink acknowledgements). From the received DL-ACK
frames, the AP uses the packet size information and applies the
proposed resource allocation algorithm. The AP sorts all N stations in
decreasing order of packet size. This is followed by grouping. Once
the grouping is done the AP allocates one sub-channel per station in
each group. The transmission time for each group is determined by the
maximum packet length in each group.
24
Figure 10: Grouping based MAC protocol procedure.
3. Data Transmission phase:
The data transmission by STAs takes place only on the allocated sub-
channels. Finally the process comes to an end when AP sends UL-
ACK (uplink acknowledgement).
3.3 Grouping based Resource Allocation
The main idea is to form groups of stations on the basis of packet lengths,
followed by resource allocation to each group. Since the grouping is carried
out after the stations have been sorted in decreasing order of their packet length,
each group contains stations having approximately the same packet length or
25
very less difference. As a result, when stations transmit data in their allocated
sub-channels, the channel wastage is less in comparison to the scenario where
the AP allocates radio-resource without considering packet length.
3.3.1 Grouping based Resource Allocation algorithm
The algorithm is given as:
Step 1. After receiving DL-ACKs from all N stations, the AP checks the
Length field.
Step 2. The stations are sorted in descending order of their packet sizes.
Step 3. The AP groups the stations, with number of groups given by
Number of Groups, n = (3)
where N is number of stations and K number of sub-channels.
Step 4. Each Station of a group is assigned 1 sub-channel for data
transmission.
Step 5. The allowed transmission time of a group is calculated by
(4)
26
3.3.2 Example of proposed algorithm
Let us consider a case where the WLAN system consists of 8 stations and 4
sub-channels. Let us assume that the stations have following uplink data to
transmit, as shown in Figure 11. On application of our proposed algorithm, the
groups will be formed as seen in Figure 11.
Figure 11: Example of grouping based resource allocation.
For example, when N = 8 and K = 4, if stations have following uplink data to
transmit with sizes of – 200, 1200, 1000, 600, 1400, 1100, 500 and 300 bytes.
Then according to the algorithm in this paper, two groups will be formed with
Group2 containing stations which have 600, 500, 300 and 200 bytes and 1400,
1200, 1100 and 1000 will be in Group 1. When this scenario is compared to
random allocation, we can see that spectrum utilization is better in our case.
27
3.4 Fragmentation based Resource Allocation
The grouping based resource allocation scheme improves throughput,
however the throughput degrades drastically due to presence of idle sub-
channel. This happens when for K sub-channels, the number of stations N in
Gth group is less than the number of groups, where K is the number of sub-
channels and G is the number of groups.
Hence, we propose a resource allocation algorithm which implements the
concept of fragmentation to overcome this drawback. When (K(G-1)+1 ≤ N ≤
(K+1)(G-1)), the AP applies the proposed resource allocation algorithm and
fragments the packet of that station. The fragments are transmitted in the idle
segment of the sub-channel, hence leading to better channel utilization. The
proposed scheme can be applied in uplink as well as downlink transmission.
Hence, we will describe the algorithm for downlink transmission. The
exchange of control messages takes place in a similar pattern to that in
grouping based resource allocation. The details of resource allocation
algorithm have been depicted in Figure12.
3.4.1 Fragmentation based Resource Allocation algorithm
Step 0. Information collection:
The AP collects packet length information from all N stations.
28
29
Figure 12: Flowchart of proposed resource allocation algorithm.
Step 1. Grouping of stations:
The AP groups stations in decreasing order of packet length. The number
of groups, G is given by . Then the AP checks if the condition, (K(G-1)+1
≤ N ≤ (K+1)(G-1)) holds. In case it is false, then fragmentation is not applied
and the AP allocates 1 sub-channel to each station. However, if the condition
is true we proceed to the next step.
Step 2. Fragmentation:
The AP calculates the duration required to transmit fragments of packet of jth
station, where j is index of the stations arranged in descending order of their
packet lengths. Initially the value of j is set to N. The required duration, Treq is
determined by
Treq = (LPj + 2 Hmac + 2 Lcrc)/R, (5)
where LPj is the length of jth packet, Hmac is the length of MAC header and Lcrc
is the length of CRC(cyclic redundancy check) in bits. R is the physical data
rate.
Next, we need to evaluate the time remaining in gth group, where g is the
index of group. We begin by assigning (G-1) to g. The duration is given by
Trem_g=TSCH#K+TSCH#(K-1) , (6)
30
where TSCH#K denotes the time remaining in SCH#K and TSCH#(K-1) is the time
remaining in SCH#(K-1). In our explanation K = 4, hence we have TSCH#3 and
TSCH#4.
There are two reasons for using the last two sub-channels for fragments
(i) To maximize the probability of finding higher remaining time. Since
the groups have stations sorted in decreasing order of their packet
lengths, the last two stations of the group will be having minimum
packet lengths. Hence, when a sub-channel is allocated to the
stations sequentially, the last two sub-channels will have maximum
idle time, which can be utilized for fragments of other packets.
(ii) To keep fragmentation overhead within a limit.
Now the AP needs to check if required time is less than the remaining time
in gth group. If the condition is false, then the AP just allocates 1 sub-channel
to each station. Otherwise, it moves to next step.
Step 3. Sub-channel allocation:
The AP repeats the previous step for (j-1)th station and (g-1)th group. This
continues until (j ≤ K(G-1)). Then the AP updates the number of groups as G
= G-1 and allocates SCH#1, SCH#2 to one station each, SCH#3 and SCH#4
to two stations, where one station can send its packet and another station can
transmit fragment of its packet.
Step 4. End of resource allocation:
With this the resource allocation algorithm comes to an end.
31
3.4.2 Example of proposed algorithm
In order to explicate our algorithm, we will be citing an example. Let us
consider the case where N = 10 and K = 4. Then if grouping algorithm is
applied there will be three groups as seen in Figure 13(a). But on application
of our proposed algorithm there will be two groups as seen in Figure 13(b).
According to the proposed algorithm, the AP will first group the stations in
(a)
(b)
Figure 13: (a)Data transmission without fragmentation (b) Data transmission with fragmentation.
decreasing order of their packet lengths. Then it will check if (K(G-1)+1 ≤ N
≤ (K+1)(G-1)), which is true in our case. Next, the AP calculates the time
32
remaining in SCH#3, SCH#4 and also estimates the time required by the
packet along with fragmentation headers. If the remaining time is much higher
than the required time, then the AP fragments the packet to Fg0, Fg1 and
allocates SCH#3, SCH#4 to the fragments. The main idea is to reduce resource
wastage. Therefore, instead of transmitting packets of 10 stations in three
groups, the same can be done in two groups. This implies maximum utilization
of channel bandwidth. Figure 13(b) illustrates the data transmission when our
proposed scheme is implemented. We can see that the fragments of packet of
STA 5 and STA 7 have been allocated to sub-channels of Group-1 and Group-
2 respectively.
3.5 Frame Structure
In order to support our protocol, we introduce some changes in frame
structures. The RTS frame, shown in Figure 14(a), is sent by AP to all N
stations. The ‘Order’ field contains the order in which stations are supposed to
reply. The CTS frame that can be seen in Figure 14(b), has no changes
introduced. The RAI frame has ‘Group no.’ field to notify assigned group
number. There is ‘SCH ID’ field that follows each ‘RA’ field to notify the
assigned SCH, and ‘Time’ field which denotes the transmission time allocated
to each group. The DL-ACK frame, as seen in Figure 14(d), is transmitted by
the stations, to inform the AP the result of transmission i.e. if it was successful
or not. The ‘Result’ field indicates success or failure. The most important field
in this frame is ‘Length’, which contains the packet lengths of all stations eager
to transmit. Using this information, the AP performs resource allocation.
33
Finally, the UL-ACK frame as seen in Figure 14(e), is used by the AP to send
acknowledgement to stations. The ‘Result’ field indicates if transmission is a
success or failure. It is set to 1 if all stations have succeeded and 0 if there is a
failure.
Figure 14:Modified Frame format.
3.6 Fragmentation
Fragmentation is the process of partitioning an MSDU(MAC service data
unit) into smaller MAC level frames, MPDUs(MAC protocol data units). Each
34
fragment shall contain a Sequence control field, which is comprised of a
sequence number and fragment number. The Sequence control field allows the
destination station to check that all incoming fragments belong to the same
MSDU, and the sequence in which the fragments should be reassembled.
When a station is transmitting an MSDU, the sequence number shall remain
the same for all fragments of that MSDU. The Frame control field also
contains a bit, the More fragments bit, that is equal to 0 to indicate the last (or
only) fragment of the MSDU. The header of each fragment contains the
Sequence control field and the More Fragments indicator. Figure15. shows the
fragmentation of MSDU into MPDUs.
Figure 15: Fragmentation of MSDU.
35
Chapter 4
Simulation Results
For performance evaluation, we have used throughput as the metric and
compared the results with existing scheme. We have assumed that the BSS
consists of 1 AP and N stations, and operates in infrastructure mode. The PHY
layer specifications are based on that defined in IEEE 802.11ac standard [29],
[30]. The packets generated by stations are uniformly distributed, and their
size is in the range of 200-1500 bytes. The MCS(modulation and coding
scheme) parameter sets are shown in Table 1.
Table 1: Modulation and Coding Scheme in 20MHz.
Modulation Coding Data Rate (Mbps)
QPSK 1/2 14.40
QPSK 3/4 21.70
16-QAM 1/2 28.90
16-QAM 3/4 43.30
64-QAM 2/3 57.80
64-QAM 3/4 65.00
64-QAM 5/6 72.20
36
Table 2: Simulation parameters.
Parameters Values
Bandwidth B 20 MHz
Number of SCHs K 4
Number of STAs N 4-16
Physical Data Rate R 65 Mbps
Basic Data Rate Rb 7.2 Mbps
Length of PHY header 120 bits
Length of MAC header 240 bits
SIFS 10 us
Δ 1 us
Modulation 64 QAM
Code Rate 3/4
MCS 6
Packet Size 200-1500 bytes
The parameters used for simulation are mentioned in Table 2.
37
For our system, we need to find the saturation throughput. It is denoted by S,
defined as the ratio of total data successfully transmitted (uplink and
downlink), to the total channel time and is derived as follows:
S =∑
(7)
where E[ ] is the expected packet length of jth station. Ttotal is the total time
taken for exchange of control messages (RTS, CTS, RAI and ACK) and
actual data transmission.
The duration of RTS and CTS exchange phase can be found by equations
(8) and (9),
TRTS= / + / + δ , (8)
TCTS= + N( + )/ + δ, (9)
where δ is propagation delay, R and Rb are operational and basic rates, i.e. 65
Mbps and 7.2 Mbps respectively. TRTS and TCTS are the time taken to transmit
RTS and CTS frames respectively. Lrts is the length of the RTS message and
Lcts is the length of CTS. Hphy is the length of the PHY header and TSIFS is the
duration of SIFS. The duration of transmitting resource allocation information
from AP to the stations is denoted as TRAI and given by:
TRAI = + + / + δ, (10)
38
where Lrai is the length of resource allocation information that the AP sends.
The duration of data transmission phase is given by equations (11) and (12),
=∑
, (11)
=∑
, (12)
here Gi denotes the ith group and max (DL/UL- ) is the maximum packet
length of all the stations in that ith group. Hmac is the length of MAC header.
From equations (13) and (14), we can calculate the time required for
exchanging ACKs. LDL-ack and LUL-ack are the lengths of downlink and uplink
acknowledgements respectively.
TDL-ACK= + N( + / + δ , (13)
TUL-ACK= + ( + / + δ . (14)
Finally the total duration, Ttotal can be calculated by:
Ttotal= TRTS +TCTS +2TRAI + + + TDL-ACK +TUL-ACK . (15)
39
5.1 Proposed Schemes vs Novel DCF
Figure 16 shows the throughput performance for increasing number of
stations, for a fixed data rate of 65Mbps. Here, the throughput is defined as
ratio of total data successfully transmitted (uplink and downlink), to the total
channel time. From Figure 16, we can deduce that, compared to novel DCF
[11] , both of our proposed algorithms perform better. The throughput of novel
DCF decreases with increasing number of stations, as amount of data
transmitted reduces due to the complexity of its algorithm. The throughput of
grouping algorithm shows best results when N = 2 K, n≥2. But we can see
throughput decreases drastically when N = 5,9,10,13…, because of presence
Figure 16: Throughput vs increasing number of stations.
40
idle sub-channels.
On the other hand, the throughput of proposed algorithm based on
fragmentation increases linearly with number of stations and shows almost 10%
improvement compared to existing novel DCF scheme. We also see that our
proposed algorithm based on fragmentation performs better than grouping
algorithm when N = 5,9,10,13,14,15. This trend can be justified by the fact
that our algorithm implements fragmentation and reduces the number of
groups. Hence, the spectral efficiency gets better. Our results show that our
proposed algorithm performs best when the number of stations is more than
the number of sub-channels (K).
5.2 Throughput vs PHY data rates
Figure 17: Throughput vs PHY data rates and number of stations (grouping).
41
To see the effect of increasing PHY data rates on network, we observed the
throughput with increasing number of stations and varying PHY data rates.
Figure.17 and Figure.18 illustrate the behavior of our proposed algorithms at
different PHY data rates. From Figure 17, we can conclude that the throughput
keeps increasing with increasing data rates. Throughput is highest when data
rate is 72.20 Mbps. Similar trend in seen in Figure 18, where we calculate the
throughput.
Figure 18: Throughput vs PHY data rates and number of stations (fragmentation).
42
5.3 Throughput vs number of sub-channels
In order to see how the throughput changes with number of SCHs, we
calculate the throughput for K = 4 and 8 and N = 8, 12, 16, 20, 24, 28, 32.
From the results seen in Figure 19, we can conclude that with increasing
number of SCHs, throughput also increases. We see that there is almost 10%
improvement in throughput when number of SCHs is increased from 4 to 8.
Figure 19: Throughput vs number of sub-channels.
43
5.4 Simulation result vs Mathematical result
We compared the analytical results with the simulation results (obtained
using MATLAB tool) for both our proposed resource allocation schemes.
From the results shown in Figure 20 and Figure 21, we can see that there is
very less difference between analytical results and mathematical results.
Therefore, we can say that our proposed protocol is accurate.
Figure 20: Simulation results vs Mathematical Results (grouping)
44
Figure 21: Simulation results vs Mathematical Results (fragmentation)
45
Chapter 5. Conclusion
In this thesis, we proposed a simple resource allocation algorithm, which
implements a combination of grouping and fragmentation to reduce radio
resource wastage and improve spectral efficiency. There are many existing
MAC protocols for OFDMA WLANs. Most of them wanted to improve
throughput by attaining multi-user diversity gain, QoS satisfaction, overhead
reduction etc but did not consider the presence of idle sub-channels. Moreover,
most of them are applicable to environment with stations having homogeneous
packet lengths. However, in real stations have heterogeneous packet lengths.
The resource allocation algorithm proposed are quite complex.
Therefore, in this thesis, we have contributed a simple resource allocation
scheme which is applicable to WLAN environment in which stations have
heterogeneous packet lengths i.e. all stations have data of different lengths in
their transmission buffer. The AP will allocate sub-channels to stations on the
basis of packet length information. First, the stations are sorted in decreasing
order of their packet lengths and grouped. If the number of stations in Gth
group is less than the number of groups, then fragmentation is implemented.
To check the effectiveness of our proposed scheme we carry out
performance evaluation. From simulation results, we found that the proposed
scheme shows enhanced throughput with increasing number of stations
compared to existing schemes. We also compared the simulation results with
analytical results to check the accuracy of our analytical model and found that
they were very close. Our work can be applied to other 802.11 standards like
802.11ax as well as future 5G networks.
46
References
1. K. Pahlavan and A.H. Levesque, “Wireless data communications,” Proc.
IEEE ,vol. 82, pp. 1398-1430, Sept. 1994.
2. A. De Simone and S. Nanda, "Wireless data: Systems, standards,
services," J. Wireless Networks, vol. 1, no. 3, pp. 241-254, Feb. 1996.
3. G. Bianchi, “Performance Analysis of the IEEE 802.11 Distributed
Coordination Function,” IEEE Journal On Selected Areas in
Communications, vol. 18, no. 3, pp. 535-547, Mar. 2000.
4. H. Yin and S.Alamouti, “OFDMA: A Broadband Wireless Access
Technology”, in Sarnoff Symposium, pp. 1-4, Mar. 2006.
5. X. Wang, W. Xiang, “An OFDM TDMA/SA MAC Protocol with QoS
Constraints for Broadband Wireless LANs”, Wireless Networks, vol.12,
pp 159-170, April 2006.
6. S. Srikanth, V. Kumaran, C. Manikandan, “Orthogonal frequency division
multiple access: Is it the multiple access system of the future?” [Online].
Available: http://aukbc.org/comm/Docs/Tutorils/OFDMA_BCW_cv6.pdf
7. Y. P.Fallah, P.Nasiopoulos and H. M. Alnuweiri, “Hybrid
OFDMA/CSMA Based Medium Access Control for Next-Generation
Wireless LANs”, IEEE International Conference on
CommunicationsICC’08, pp. 2762-2768, May 2008.
8. S. Valentin, T. Freitagand H. Karl, “Integrating multiuser dynamic
OFDMA into IEEE 802.11 WLANs – LLC/MAC extensions and system
performance”, IEEE International Conference on Communications
ICC’08, pp 3328-3334, May 2008.
47
9. H. M. Alnuweiri, Y. P. Fallah and P. Nasiopoulos, “OFDMA-Based
Medium Access Control for Next-Generation WLANs”, EURASIP
Journal on Wireless Communications and Networking, vol. 2009, pp. 1-9.
Feb. 2009.
10. G. Haile, J. Lim, “C-OFDMA: Improved Throughput for Next Generation
WLAN Systems Based on OFDMA and CSMA/CA”, 4thInternational
Conference on Intelligent Systems, Modelling and Simulation, IEEE, 2013,
pp. 497-502, Jan. 2013.
11. S. Miyamoto, S. Sampei and W. Jiang, “Novel DCF-based multi-user
MAC protocol for centralized radio resource management in OFDMA
WLAN systems”, IEICE Transactions on communications, vol.E96-B,
no.9 pp. 2301-2312, Sept. 2013.
12. K. Shimamoto, S. Miyamoto, S. Sampei, W. Jiang, “Two-stage DCF-
based access scheme for throughput enhancement of OFDMA WLAN
systems,” Wireless Personal Multimedia Communications (WPMC), 2012
15th International Symposium on, pp. 584-588, Sept. 2012.
13. K. Tan, J. Fang,Y. Zhang, S.Chen, L. Shi, J. Zhang, Y. Zhang,“Fine-
grained channel access in wireless LAN,” SIGCOMM '10, pp 147-158,
Sept. 2010.
14. X. Wang, H. Wang,“A Novel Random Access Mechanism for OFDMA
Wireless Networks,” Global Communications Conference (GLOBECOM
2010), 2010 IEEE, Dec. 2010.
15. E. Yaacoub, Z. Dawy, “A Survey on Uplink Resource Allocation in
OFDMA Wireless Networks,” IEEE Communications Surveys &
Tutorials, vol. 14, pp. 322-337, May 2011.
48
16. B. Li, Q. Qu, Z. Yan, M. Yang, “Survey on OFDMA based MAC protocols
for the next generation WLAN”, Wireless Communications and
Networking Conference Workshops (WCNCW), 2015 IEEE, pp 131-135 ,
Mar. 2015.
17. Q. Qu, B. Li, M. Yang and Z. Yan, “An OFDMA based concurrent
multiuser MAC for upcoming IEEE 802.11ax”, Wireless Communications
and Networking Conference Workshops (WCNCW), pp. 136-141, Mar.
2015.
18. H. Zhou , B. Li , Z. Yan , M. Yang , Q. Qu, “An OFDMA based multiple
access protocol with QoS guarantee for next generation WLAN”, Signal
Processing, Communications and Computing (ICSPCC), 2015 IEEE
International Conference on, Sept. 2015.
19. M. Mohanty, J. K. Kim, J. H. Kim, “Grouping-Based Resource Allocation
Scheme for Spectral Efficiency Enhancement of OFDMA WLAN
Systems”, The Korean Institute of Communications and Information
Sciences, Jeju Island, 2016, pp. 210-211, June 2016.
20. A. Jamalipour, T. Wada, and T. Yamazato, “A tutorial on multiple access
technologies for beyond 3G mobile networks,” IEEE Communications
Magazine, vol. 43, no. 2, pp. 110–117, 2005.
21. H. Kwon and H. Seo and S. Kim and B. G. Lee, “Generalized CSMA/CA
for OFDMA systems: protocol design, throughput analysis, and
implementation issues,” IEEE Transactions on Wireless Communications,
vol. 8, no. 8, pp. 4176–4187, Aug. 2009.
49
22. H. Kwon and H. Seo and S. Kim and B. G. Lee, “Opportunistic multi-
channel CSMA protocol for OFDMA systems,” IEEE Transactions on
Wireless Communications, vol. 9, no. 5, pp. 1552- 1557, May 2010.
23. J. Jung and J. Lim, “Group Contention-Based OFDMA MAC Protocol for
Multiple Access Interference-Free in WLAN Systems,” IEEE
Transactions on Wireless Communications, vol. 11, no. 2, pp. 648-658,
Feb. 2012.
24. Y. J. Choi, S. Park, and S. Bahk, “Multichannel random access in OFDMA
wireless networks,” IEEE Journal on Selected Areas in Communications,
vol. 24, no. 3, pp. 603–613, 2006.
25. F. Cuomo, A. Baiocchi, and R. Cautelier, “A MAC protocol for a wireless
LAN based on OFDM-CDMA,” IEEE Communications Magazine, vol. 38,
no. 9, pp. 152–159, 2000.
26. T. Mishima, S. Miyamoto, S. Sampei, W. Jiang, “Novel DCF-based multi-
user MAC protocol and dynamic resource allocation for OFDMA WLAN
systems”, 2013 International Conference on Computing, Networking and
Communications (ICNC), pp. 616-620, Jan. 2013.
27. J. Mo, H. S. So, and J. Walrand, “Comparison of multichannel MAC
protocols”, IEEE Transactions on Mobile Computing, vol. 7, no. 1, pp. 50–
65, Jan. 2008.
28. H. Ferdous and M. Murshed, “Enhanced IEEE 802.11 by integrating
multiuser dynamic OFDMA,” Wireless Telecommunications Symposium
(WTS), pp. 1–6, April 2010.
50
29. 802.11ac In-Depth - Aruba Networks, [Online]. Available:
http://www.arubanetworks.com/pdf/technology/whitepapers/WP_80211acIn
Depth.pdf
30. WLAN 802.11ac Technology white paper,[Online]. Available:
http://e.huawei.com/sg/marketingmaterial/onlineview?materialid=%7B2f
4be832-3aeb-41a0-9bab-766702da8d9b%7D