Content uploaded by Ryo Hamamoto
Author content
All content in this area was uploaded by Ryo Hamamoto on Feb 12, 2018
Content may be subject to copyright.
Performance Evaluation of Control Method to
Guarantee Throughput based on Media Access
Control SP-MAC over WLAN
Yuuki Yamamoto, Ryo Hamamoto, Hiroyasu Obata, Chisa Takano, and Kenji Ishida
Graduate School of Information Sciences, Hiroshima City University,
3-4-1 Ozuka-Higashi, Asa-Minami-Ku, Hiroshima, 731-3194 Japan
Email: {yamamoto@net.info., ryo@net.info., obata@, takano@, ishida@}hiroshima-cu.ac.jp
Abstract—Wireless local area networks (WLANs) based on
IEEE802.11 usually uses CSMA/CA for a media access control. In
the CSMA/CA, it is known that data frame collisions often occur
if the number of wireless terminals increases. This issue causes
the degradation of the total throughput of terminals connected
to the same Access Point (AP). To solve this problem, a media
access control method based on synchronization phenomena
of coupled oscillators (SP-MAC) has been proposed. SP-MAC
can avoid data frame collisions drastically and improve the
throughput compared with CSMA/CA. Additionally, we have
proposed a control method to guarantee throughput based on
SP-MAC. However, because the previous study evaluated only
the basic performance, the evaluation of the control method
was not sufficient. Thus, this study evaluates the control method
to guarantee throughput based on SP-MAC considering the
propagation delay. From the simulation results, we found that
the control method can secure the throughput if the propagation
delay is less than 100 ms.
Index Terms—WLAN, Media access control, Guaranteeing
throughput, Synchronization phenomena of Coupled oscillators
I. Introduction
Because wireless Local Area Networks (WLANs) based
on IEEE802.11 standard [1] have become popular, the num-
ber of WLAN users is increasing rapidly. WLAN based on
IEEE802.11 uses Carrier Sense Multiple Access with Collision
Avoidance (CSMA/CA) for a Media Access Control (MAC).
In CSMA/CA, however, a total throughput of all wireless
terminals connected to the same Access Point (AP) drastically
decreases when a lot of users connect the same AP [2]. This
is because that the number of data frame collisions increases
in this situation. To address this issue, we have proposed a
media access control method based on the synchronization
phenomena of coupled scillators (SP-MAC) [3]. SP-MAC is an
extension of CSMA/CA and it calculates a back-offtime using
the synchronized phase with phase shifting based on the syn-
chronization phenomena of coupled oscillators obtained from
Kuramoto model [4] instead of using the random integer in
CSMA/CA. The previous evaluation [5] shows that SP-MAC
can decrease data frame collisions drastically and improve the
total throughput of all terminals compared with CSMA/CA.
Moreover, [5] shows that SP-MAC can solve the unfairness
problem of throughput [6] among TCP flows.
Here in the current Internet services, a lot of users have
portable wireless devices such as smart phones and tablet PCs.
In this situation, the streaming applications such as video chat
and video streaming are often used. However, because the
bandwidth of WLAN usually dynamically changes, a QoS con-
trol for WLAN has become one of important issues. To solve
this issue, various QoS control methods [7]–[10] for WLAN
environment have been proposed. Note that this paper focuses
on the throughput as a QoS metric same as previous studies
[7]–[10]. The previous QoS control methods decreases the
total throughput by collisions when the number of terminals
increases. On the other hand, because SP-MAC can drastically
avoid collisions, we assumed that a QoS control based on SP-
MAC can obtain higher total throughput. Therefore, we have
proposed a control method to guarantee throughput based on
SP-MAC [11]. The control method uses the target throughput
required by applications and the number of total terminals
connected to the AP in order to secure the throughput. Then,
we have confirmed that the control method can guarantee the
terminal’s throughput without decreasing the total throughput.
Here the previous study [11] only evaluates the performance
over the environment when the number of terminals changes.
However, in the real environment, the propagation delay also
changes according to a path between the sender and receiver.
Especially, because TCP needs to wait for acknowledgements
from the receiver, it is assumed that securing the throughput
becomes difficult when the propagation delay increases. Thus,
this study evaluates the throughput guarantee method based
on SP-MAC considering the propagation delay.
The rest of this paper is organized as follows. Sect.II
explains related works of this paper. Next, Sect.III and Sect.IV
explain SP-MAC and the throughput guarantee method based
on SP-MAC. Then we evaluate the throughput guarantee
method considering propagation delay in Sect.V. Finally,
Sect.VI summarizes this paper.
2015 Third International Symposium on Computing and Networking
2379-1896/15 $31.00 © 2015 IEEE
DOI 10.1109/CANDAR.2015.80
265
II. Related works
This section explains CSMA/CA and the synchronization
phenomena of coupled oscillators.
A. CSMA/CA
In IEEE 802.11 WLAN networks, a wireless terminal uses
CSMA/CA as MAC and sends data frames autonomously.
Thus, each wireless terminal determines data transmission
timing individually. In CSMA/CA, if the channel becomes idle
when a data frame arrives in the transmission queue, it defers
to DCF inter frame space (DIFS) time. Then, if the channel
remains idle after DIFS, CSMA/CA waits for the back-off
time, which is randomly calculated using a contention window
(CW). Subsequently, if the channel remains idle after the back-
offtime, the terminal sends the data frame. The back-offtime
is determined using Eq.(1), which is independently calculated
by each terminal.
Backoff=Random() ×SlotTime (1)
In Eq.(1), Random() and SlotTime indicate a random integer
derived from a discrete uniform distribution [0,CW] and the
slot time interval specified in IEEE 802.11, respectively. At
this point, the initial CW is set to CWmin. If a collision causes
the data frame transmission to fail, then the terminal again
sets the back-offtime using Eq.(1). In this case, the CW
becomes twice the previous value, and the upper bound is
CWmax. If the retransmission exceeds the maximum retry limit
(usually seven), the terminal discards the data frame. Here in
CSMA/CA, all wireless terminals have equal priority.
B. Synchronization phenomena of coupled oscillators
Synchronization indicates that the phenomena caused by
multiple oscillators with different periods transform incoherent
rhythms into synchronized rhythms with each interaction. This
is also observed in nature, such as the synchronous flashing of
fireflies [12] and synchronization of metronomes [13]. These
synchronized oscillators are called coupled oscillators. During
synchronization, the phase differences and frequencies of all
coupled oscillators converge at certain values.
One of typical models of synchronization phenomena is the
Kuramoto model [4]. SP-MAC uses the synchronization of N
coupled oscillators of the Kuramoto model. In the Kuramoto
model, the i-th oscillator runs independently at its own natural
frequency ωiand interacts with all others. Then, the i-th
oscillator’s phase θi(0 <θ≤2π) is calculated using Eq.(2).
dθi
dt=ωi+K
N
N
j=1
sin(θj−θi)(i=1,2,··· ,N) (2)
In Eq.(2), K(>0) indicates coupling strength. The second term
is an interaction term, which is standardized by K/Nto be
independent of system size N.
III. SP-MAC: Media access control method based on the
synchronization phenomena of coupled oscillators
SP-MAC uses the synchronized phase with phase shifting
(see Fig.1) based on Eq. (2) to set the back-offtime instead of
Fig. 1. Time change of synchronized cosine value with phase shifting (the
number of oscillators Nis 2).
using a random integer, thereby avoiding overlap of the back-
offtime among terminals. Furthermore, SP-MAC is based on
the original CSMA/CA (i.e., only the calculation of the back-
offtime is different); thus, it can be used in an environment in
which SP-MAC and CSMA/CA terminals coexist. Note that
the SP-MAC method sets the following preconditions.
•The number of wireless terminals does not change after
data transmission has begun.
•The AP and all wireless terminals do not move.
•The AP and all wireless terminals employ SP-MAC.
•The AP and all wireless terminals do not use the
RTS /CTS function.
In SP-MAC, an AP determines the natural frequency ωiand
coupling strength Kthat satisfy the synchronizing condition
according to Nprior to starting transmission (iis the ID).
To satisfy the condition that each oscillator synchronizes with
phase shifting, SP-MAC adopts a different ωifor each wireless
terminal (i.e., no overlap occurs among all ωi). Next, the AP
sets an ID i(1 ≤i≤N) for each wireless terminal and applies
ωiand an initial phase θi(0) to the i-th wireless terminal. Each
initial phase θi(0) has a different value to avoid collision at
the beginning of the data transmission. Then, using a beacon,
the AP sends the control parameters i,θi(0), ωi,K, a control
interval Δt, and Nfor all wireless terminals.
After receiving the beacon, each wireless terminal im-
mediately begins calculation of the phase using the control
parameters. Next, the wireless terminal calculates the phase
θi(t) for all ID iusing Eq.(2) for every Δt. The calculation of
the phase continues while the terminal connects to the AP even
if no data exists for transmission. When the wireless terminal
wants to send a data frame at time t, it calculates the back-
offtime (Backoff) using Eq.(3) and phase θi(0) for each ID i.
Then, the wireless terminal sends the data frame in the same
manner as CSMA/CA.
Backoff=((|cos θi(t)|×α) mod N)×SlotTime (3)
In Eq.(3), slot time and αshow the slot time interval specified
in IEEE 802.11 and a coefficient for obtaining the normalized
phase, respectively. In this study, we set αequal to 100 [3],
[5]. If the wireless terminal detects data frame collisions, it
266
Fig. 2. Communication timing when both terminals use same amplitude
(original SP-MAC).
Fig. 3. Communication timing when each terminal uses different amplitude
(throughput control method).
calculates the new back-offtime using Eq.(3) and the phase
when a collision is detected again.
IV. Control method to guarantee throughput based on
SP-MAC
This section explains the control method to guarantee
throughput based on SP-MAC which is proposed by [11]. We
refer to this control method based on SP-MAC as a throughput
control method in this paper.
In order to control the throughput, the throughput control
method changes the range of the back-offtime of SP-MAC
(the amplitude of the cosine wave) using the method in [14].
In [14], the range of back-offtime (amplitude) of AP is only
changed. On the other hand, the throughput control method
changes the range of back-offtime for each terminal. For
example, Fig.2 and Fig.3 show the cosine values and com-
munication timings of two terminals. Fig.2 and Fig.3 indicate
the case when both terminals use same amplitude (original SP-
MAC) and each terminal uses different amplitude (throughput
control method), respectively. In these figures, each arrowed
line shows the communication timing when each terminal
can send a data frame because the back-offtime depends on
the cosine value. From Fig.2, if both terminals have same
amplitude (original SP-MAC), each terminal can obtain fair
communication timing. In contrast, from Fig.3, if terminal
A has smaller amplitude than terminal B (the throughput
control method), terminal A can obtain communication timing
frequently compared with terminal B.
!
!
Fig. 4. Simulation model.
Thus, the throughput control method can control the
throughput of each terminal by using amplitude. Therefore,
in the throughput control method, the back-offis obtained by
Eq.(4).
Backoff=Ai×((|cos θi(t)|×α) mod N)×SlotTime (4)
In Eq.(4), Aiindicates the amplitude. If Aiis less than 1, the
terminal can send a data frame preferentially than a terminal
with Ai≥1. In this study, we refer to the terminal with shorter
Ai(≤1) as a priority terminal. Also, we refer to the other
terminal (Ai=1) as a non-priority terminal. Furthermore,
in order to set appropriate Aifor the target throughput Thr
required from applications, the throughput control method
calculates Aiby using the number of total terminals connected
to the AP NTand Thr. Note that this equation is obtained by
the preliminary evaluations in [11].
Ai=Thr ·NT0.8
16.0α
(5)
α=⎧
⎪
⎪
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎪
⎪
⎩
−10.0
log(41.0NT)(UDP)
−10.0
log(22.0NT)(CUBIC-TCP)
Thus, Eq.(5) is used for setting Aiof the priority terminal.
V. S imulation experiments
This section evaluates the throughput control method con-
sidering propagation delay by using network simulator NS2
[15].
A. Simulation settings
Figure 4 shows a network model for the simulation. In
this model, IEEE802.11g (PHY) is used for the WLAN
environment. The number of priority terminals (Ai<1) and
non-priority terminals (Ai=1) are 1 and NT−1, respectively.
Moreover, all wireless terminal implement SP-MAC and each
wireless terminal (Sender) generate 180 seconds of traffic
(each wireless terminal generates one flow). In addition, the
propagation delay between AP and Router DAR is set equal to
{10,40,100,150,200}ms. This study uses UDP and CUBIC-
TCP (the standard TCP for Linux and Android OS) [16] for the
transport protocol and the segment size is 1000 Byte. More-
over, CBR (30 Mbps) and FTP are used for the application
of UDP and TCP, respectively. The buffer size of AP and all
wireless terminal are 500 and 250 packets, respectively.
267
Fig. 5. Target throughput vs. achievement ratio (UDP, NT=20).
For SP-MAC, we set the control parameters as follows
by considering the synchronization condition. First, the ini-
tial phase θi(0) and natural frequency ωiwere set to non-
overlapped values in the ranges (0,1) and [0,2], respectively.
Then, we set the coupling strength Kto 5.0. Also, we set the
parameter Nin Eq.(3) and Eq.(4) equal to 100. The simulation
results showed the averages obtained from 5 trials. In the
evaluations, the number of wireless terminals does not change
and all wireless terminal do not move.
B. Simulation results
To begin with, we show the results of UDP when NT=20.
Figure 5 shows the relationship between the target throughput
and the achievement ratio of the priority terminal for each DAR.
Note that the achievement ratio is obtained by the throughput
of priority terminal and the target throughput (Eq.(6)).
Achievement Ratio =priority terminal’s throughput
Thr ×100[%]
(6)
From Fig. 5, the throughput control method can guarantee the
priority terminal’s throughput even if DAR becomes larger in
case of UDP.
Next, we discuss the result of CUBIC-TCP. Figure 6 de-
scribes the relationship between the target throughput and
the achievement ratio of the priority terminal for each DAR.
From Fig. 6, the throughput control method cannot guarantee
the throughput of the priority terminal when DAR increases.
Thus, when CUBIC-TCP is used for the transport protocol,
the performance of the throughput control method is deeply
affected by DAR. This result is caused by the window control
of TCP. Here Figs. 7 and 8 show the time change of the
window size of the priority terminal and non-priority terminals
when DAR is 10 ms and 100 ms, respectively. From Figs. 7
and 8, the window size of the priority terminal increases
quickly compared to the ones of the non-priority terminals.
However, the time to reach the upper limit becomes longer
if DAR changes from 10 ms to 100 ms. As a result, the
Fig. 6. Target throughput vs. achievement ratio (CUBIC-TCP, NT=20).
Fig. 7. Time change of window size of the priority terminal and non-priority
terminals (NT=20, DAR =10 ms).
throughput decreases when the propagation delay increases.
Here Bandwidth Delay Product (BDP) is known as a suitable
upper limit of window size. BDP is obtained by Eq.(7). In
Eq.(7), TRand Bshow the round trip time between sender and
receiver and the bandwidth of the bottleneck link, respectively.
W=TR×B
packet size (7)
In Fig. 6, however, the upper limit of window size is fixed
value 128 (standard for Linux OS) regardless of DAR.Asa
result, the throughput control method cannot guarantee the
throughput of the priority terminal. Therefore, in the next
evaluation, we use BDP for the upper limit of window size.
Figure 9 shows the achievement ratio of priority terminal’s
throughput for each DAR (CUBIC-TCP) when the upper limit
of window size is set equal to BDP. From Fig. 9, the
achievement ratio improves compared with the one in Fig. 6
by modifying the upper limit of window size. However, the
achievement ratio decreases when DAR >100 ms. Here Fig. 10
and Fig. 11 represent the time change of the window size
of the priority terminal and non-priority terminals when the
upper limit of window size is set equal to BDP. Figure 10 and
Fig. 11 show the result of DAR =10 ms and DAR =100 ms,
respectively. From Fig. 10, the window size of priority terminal
268
Fig. 8. Time change of window size of the priority terminal and non-priority
terminals (NT=20, DAR =100 ms).
Fig. 9. Target throughput vs. achievement ratio (CUBIC-TCP, NT=20, the
upper limit of window size is set equal to BDP).
has almost stable. On the other hand, from Fig. 11, the window
size drastically changes when the propagation delay increases.
Here BDP of DAR =10 ms and DAR =100 ms are 193 packets
and 1066 packets, respectively. On the other hand, the buffer
size of the wireless terminal is 250 packets. Thus, BDP is
greater than the buffer size when DAR is 100 ms. As a result,
the buffer overflow occurs in MAC layer of wireless terminals
because a lot of data segments are sent from TCP layer. Then,
the window size decreases by congestion control.
From above resuls, we found that the unstable window size
is caused by buffer overflow. To address this issue, we set
all wireless terminal’s buffer size equal to 2×BDP. Figure 12
shows the achievement ratio of priority terminal’s throughput
for each DAR (CUBIC-TCP) when the upper limit of window
size and buffer size are set equal to BDP and 2×BDP,
respectively. From Fig. 12, if both the upper limit of window
size and the buffer size of the sender are modified when
DAR ≤100 ms, the throughput control method can guarantee
the priority terminal’s throughput. However, if DAR >100 ms,
it cannot secure the throughput. This is because that the
waiting time of TCP-ACK becomes larger by increasing DAR.
As a result, the priority terminal cannnot obtain the sufficient
Fig. 10. Time change of window size for the priority terminal and non-priority
terminals (NT=20, DAR =10 ms, the upper limit of window size is set equal
to BDP).
Fig. 11. Time change of window size for the priority terminal and non-priority
terminals (NT=20, DAR =100 ms, the upper limit of window size is set
equal to BDP).
throughput. Here, Fig. 13 plots the time change of the priority
terminal’s throughput for each DAR when the target throughput
Thr is 15 Mbps. From Fig. 13, when DAR is 10 ms and 100 ms,
it takes about 4 sec and 30 sec until the throughput exceeds
Thr, respectively. That is, the time when the window size
increases the sufficient size to satisfy Thr increases as DAR
increases. As a result, the throughput control method cannot
obtain Thr.
From simulation results, when the propagation delay is
less than 100 ms (e.g. between Japan and U.S.A. [17]) in
case of TCP, the throughput control method can guarantee
Thr. However, if the propagation delay becomes larger than
100 ms (e.g. satellite communication system), it is difficult to
guarantee the throughput of the priority terminal. In order to
obtain sufficient throughput in long deay environment, we have
to consider new TCP congestion control for the throughput
guarantee method.
VI. Conclusion
In this study, we evaluated a control method to guarantee
throughput based on SP-MAC considering the propagation
delay. As a result, we found that the throughput control method
can control the terminal ’s throughput even if the propagation
delay changes when UDP is used for the transport protocol. On
the other hand, when TCP is used for the transport protocol,
269
Fig. 12. Target throughput vs. achievement ratio (CUBIC-TCP, NT=20, the
upper limit of window size and buffer size are set equal to BDP and 2×BDP,
respectively).
Fig. 13. Time change of throughput for each propagation delay (NT=20,
Thr =15 Mbps, the upper limit of window size and buffer size are set equal
to BDP and 2×BDP, respectively).
the throughput control method can secure the throughput if
the propagation delay is less than 100 ms.
The followings could be studied in future:
•Evaluation of the proposal in other network environments
and comparative evaluation using other QoS mechanisms
such as IEEE802.11e
•Consider multiple priority control such as EDCA of
IEEE802.11e
Acknowledgments
This work was partly supported by JSPS KAKENHI Grant
Numbers 26420367, 26280032, 15K00431, 15H02688, and
Project Research Grants of the Graduate School of Information
Sciences, Hiroshima City University.
References
[1] IEEE Standard, “Wireless LAN medium access control (MAC) and
physical layer (PHY) specifications,” ANSI/IEEE Std 802.11, 2012.
[2] G. Bianchi, “Performance analysis of the IEEE 802.11 distributed
coordination function,” IEEE J. Sel. Areas Commun., 18, 3, pp.535–547,
2000.
[3] R. Hamamoto, H. Obata, C. Takano, K. Ishida, and K. Yamada,
“Media access control mechanism based on synchronization of coupled
oscillators over wireless LAN,” IEICE Technical Report, IN2013-157,
2014.
[4] Y. Kuramoto, Chemical oscillations, waves, and turbulence, Springer,
1984.
[5] H. Obata, R. Hamamoto, C. Takano, and K. Ishida, “Throughput
characteristics evaluation of media access control method based on
synchronization phenomena of coupled osciators over WLAN coexisting
of CSMA/CA terminals,”Proc. the 7th International Workshop on ASON
2014, pp.254–259, 2014.
[6] S. Pilosof, R. Ramjee, D. Raz, Y. Shavitt, and P. Sinha, “Understanding
TCP fairness over wireless LAN,” Proc. IEEE INFOCOM 2003,2,
pp.863–872, 2003.
[7] J.N.A. -Karaki and J.M. Chang, “A simple distributed access control
scheme for supporting QoS in IEEE 802.11 wireless LANs,” Proc. IEEE
WCNC 2004, 1, pp.213–218, 2004.
[8] Z. Zhou, Y. Xiao, D. Wang, M.H. Lee “MAC frame control on wireless
LAN to guarantee QoS of high priority traffics,” Proc. the 5th ICWMMN
2013, pp.84–88, 2013.
[9] H. Obata, A. Momota, J. Funasaka, and K. Ishida, “A control method
of guaranteeing throughput for TCP communication considering han-
dover over WLAN,” Proc. IEEE the 13th International Workshop on
ADSN2014, pp.113–118, 2014.
[10] T. Murase, K. Uchiyama, Y. Hirano, S. Shioda, and S. Sakata, “MAC-
Frame Receiving-Opportunity Control for Flow QoS in Wireless LANs,”
IEICE Trans. Commun., E92-B, 1, pp.102–113, 2009.
[11] Y. Yamamoto, R. Hamamoto, H. Obata, C. Takano, and K. Ishida,
“Control Method to guarantee throughput based on media access control
method SP-MAC over WLAN,” IEICE technical report, NS2014-248,
2015.
[12] J. Buck, “Synchronous rhythmic flashing of fireflies. II.,” Q. Rev. Biol.,
63, 3, pp.265–289, 1988.
[13] J. Pantaleone, “Synchronization of metronomes,”Am. J. Phys., 70, 10,
pp.992–1000, 2002.
[14] R. Ando, R. Hamamoto, H. Obata, C. Takano, and K. Ishida, “An
access point control based on media access control method SP-MAC
over WLAN,” IEICE technical report, NS2014-249, 2015.
[15] Network Simulator - ns (version 2), available at http://www.isi.edu/
nsnam/ns/, 2015.
[16] S. Ha, I. Rhee, and L. Xu, “CUBIC: A new TCP-friendly high-speed
TCP variant,” ACM SIGOPS Operating System Review, 42, 5, pp.64–74,
2008.
[17] C. Marcondes, A. Persson, M. Sanadidi, M. Gerla, H. Shimonishi,
T. Hama, and T. Murase, “Inline path characteristic estimation to
improve TCP performance in high bandwidth-delay networks,” Proc.
PFLDnet 2006, 5pages, 2006.
270