Upload
urfriendlyjoe
View
214
Download
0
Embed Size (px)
Citation preview
8/10/2019 QoE-driven Wi-Fi Selection Mechanism for Next Generation Smartphones
1/6
QoE-driven Wi-Fi Selection Mechanism for NextGeneration Smartphones
Hyun Soon Kim, Eugene Kim, Hwangnam KimSchool of Electrical Engineering, Korea University
Seoul, Republic of Korea 136075Email:{gustns2010, ekim57, hnkim}@korea.ac.kr
AbstractSmartphones and portable devices are widely usedthese days. Data transmission networks such as 3G and Wi-Fiare used for most smartphones, and LTE is quickly increasingits market share in order to support enormous network userswho cause data congestion on every part of network topology.Especially, wireless part of network suffers from starvation andcongestion due to many users who wish high quality Internetservice such as HD video streaming. In this paper, we propose anefficient mechanism of auto-connecting qualified Wi-Fi network forthe next generation smartphones in order to provide acceptableInternet service for device users. One of authentication-free access
points and others that are provided by contracted Internet serviceprovider can be automatically connected with the mechanism. Also,quality of experience (QoE) of the connected AP is guaranteed byestimating Signal to Interference and Noise Ratio (SINR) whichis derived by the number of stations sharing the same channel.We implemented the proposed mechanism in both Android phoneand also OPNET simulation, and also we conducted a performanceevaluation study. The simulation results indicate that the proposedmechanism retains the best set of candidate APs to be connectedand improves the MAC layer throughput.
I. INTRODUCTION
Desire of high quality wireless network service is unex-
pectedly arising due to the boom of smartphones and Tablet
PCs. Large screen size compared to feature phones results inthe demand of high quality video or picture so that screen
resolution is always an update issue for new models of those
hand-held devices. High quality screen resolution in turn causesthe necessity of qualified network service that supports high
definition video streaming. Many Internet service providers
strive to build up APs to cover crowded zones with too many
users (which is called offloading) and develop new generation ofwireless networks such as WiMAX, LTE and WiBro to achieve
bursting user requirements. SK Telecom and LG U+, telecom
companies in Korea, already have launched LTE data serviceon the year 2011. Huge amount of investment has been already
made to commercialize LTE networks in Korea [1].
The use of wireless Internet service is not restricted to beperformed only at cellular networks but also other networks
such as WiMAX and Wi-Fi. Telecom companies in South
Korea have already constructed millions of Wi-Fi APs [2] toserve the customers who want to use qualified Internet service.
Indeed, most of hot-spots around cities perform much better
than original 3G network service in terms of transmission speed.However, a complete auto-configuration for switching back and
forth between Wi-Fi and Cellular Network and even between
APs is not supported yet, so that the switching is done manually
by device users.
On the other hand, IEEE 802.21 protocol [3] specifies media-independent handover service to provide skeleton of inter-
networking to Internet service providers. Also, it motivates
many researchers to focus on inter-networking of cellular andISM band network to provide better wireless Internet service
to service users [4]. Inter-networking and vertical handover
of different networks have been already structured, designedand proposed in perspective of the network layer. However
all those techniques assume the AP and STA to be connected
automatically in perspective of the link layer. In addition, each
network region is covered by multiple service providers and theydo not collaborate to support all the customers. A customer Cthat uses network of a service providerSshould use APs that areprovided by the service providerS, so that the Internet service isnot guaranteed to be used anywhere. Some places might have no
AP belonging to the service provider S. On the contrary, someplaces might have APs that can support the Internet service tothe customer C but there are too many customers served bythose APs, so that the service quality cannot be guaranteed
Many smartphones connects to an AP which is recorded on the
connection history database. However, the way of automaticallyconnecting AP usually results in bad Wi-Fi performance or
the failure of service continuity, which consequently turns into
using expensive cellular network without any notice to usersAlso, unnecessary handover delay and unsatisfactory QoE of
the Internet service may occur by connecting to those bad APs.
Based on these observations, we propose an efficient mecha-nism of auto-connecting a mobile device to the best Wi-Fi AP,
which guarantee the connected AP is much more qualified in
serving the Internet services than an AP that is determined bythe current connection mechanism. The proposed mechanism
consists of two schemes where the one is to discover the best
set of AP candidates for the Internet connectivity and the other
is to estimate the link quality of each AP candidate so as tochoose the best qualified AP. We implemented the proposed
mechanism in Android phone and also constructed a OPNET
simulation environment to carry out a performance evaluationstudy. The results present that the proposed mechanism can
increase the availability of network connectivity and also im-
prove the MAC throughput via qualified APs. We vision thathe proposed QoE-driven AP selection mechanism contributes
to better vertical handover solutions and the next generation
smartphones will adopt it to support complete inter-networkingamong heterogeneous networks.
The remainder of the paper is organized as follows. We
reveal default Wi-Fi connection problems in section II, and
then propose a new QoE driven AP selection mechanism in
2012 The First IEEE Workshop on Enabling Technologies for Smartphone and Internet of Things (ETSIoT)
978-1-4673-2557-8/12/$31.00 2012 IEEE 13
8/10/2019 QoE-driven Wi-Fi Selection Mechanism for Next Generation Smartphones
2/6
2
section III. We present a performance evaluation results in
section IV, and finally conclude the paper with section V.
I I . DEFAULTW I-FI C ONNECTION P ROBLEMS
In this section, we briefly reveal the main drawback ofthe current AP connection in mobile devices and analyze the
underlying causes for the drawback.
A. Basic Connection Procedure
The basic procedure of Wi-Fi Connection in current mobiledevices is composed of three steps: scanning, authentication
and association. Of course, roaming is an ideal approach to
re-associate seamlessly, and so many Internet service providersare continuing to deploy this roaming capability to their APs.
Note that the ultimate ideal approach is to unify all the APs
deployed from different service providers to make devices move
around over any network freely. However, this will not come innear future because expenses and benefits for service providers
should be considered appropriately. Therefore, a complete new
connection (disconnection due to signal loss and then connec-tion trial after few seconds of continuous disconnection state)
is the basic connection procedure in these days. Smartphonesuse the following procedure of connection:
The antenna enables its receiving mode to sense beaconmessage from any AP surrounding the device;
If the antenna senses no beacon message during a time
period (usually 100 ms), it switches to another channel torepeat the process;
If the antenna receives some beacon, the physical later
checks whether the configuration of the AP is found inthe device;
If there is no history of connection which matches with the
AP, the current scanning continuously or open next beacon
message to repeat the process;
If the configuration history confirms that the AP is con-nectable, do authentication and association to complete
connection procedure;
Scanning usually takes about 300ms to a few seconds untilit finds an appropriate beacon to use for connection. At least
100msfor each completely non-overlapping channel is scannedbecause each AP usually sends its beacon frame with theperiod of 100ms. In addition to this default connection scenario,
commercialized devices use a configuration history, which saves
all information necessary to re-connect the same Service SetID (SSID), to see whether or not the received beacon contains
SSID that has been recorded in that history. Service providers
usually use the configuration history to make users automati-
cally connected to their own APs for convenience. The processof this connection scenario is optimally fast since the scanning
time is optimized. If any beacon is received and confirmed to
contain the pre-specified connectable AP information, the deviceconnects to the AP right away without carrying out any further
procedure to see which APs are available for the user. This
procedure is optimal in terms of connection delay, but yieldsserious problems described in next subsection.
Note that the above connection procedure is implemented
in many simulation engines to mimic the existing wireless
devices, and some mobile devices and simulation engine (such
Fig. 1: Partial channel search problem
as OPNET) simply use the firstly arrived beacon message forthe connection.
B. Problem Analysis
There are three problems of the aforementioned original
Wi-Fi connection procedure. First, the partial channel search
sometimes incurs serious lack of service quality problem shownin Fig. 1. The figure shows that experimental device is connected
to myLGnet. However, AndroidHotspot1597 has much better
signal strength because of the short distance which makes highprobability of the better service if the APs functionalities are
assumed to be equal to each other. This unreasonable connection
is caused by a partial search, which means the channel search isnot completely done for the rest of available channels after the
connection is established with the first beacon message. Even if
there exists an AP that has better quality compared to others, the
device does not connect to that AP because the corresponding
to beacon frame has not arrived first.The second problem is that the device never tries to connect
to an AP that requires no authentication process, so called
authentication-free AP. This AP can be connected by any device,
and therefore the configuration history is unnecessary for theconnection process when authentication-free APs are available
However, since IEEE 802.11 protocol [5] does not specify the
way of connecting authentication-free AP, many devices are notcapable of connecting to such open APs. This leads to reducing
the number of choices for connecting to the Internet, and thus,
some places which are not covered by APs from contracted
service providers let users to manually search for an AP toconnect.
The last problem is that there is no specific service levelmetric at the moment of AP connection. Most of applications
that aid the connection of Wi-Fi define the quality of AP
in the increasing order of RSSI strength in terms of dBm[6]. It is true that the RSSI value goes higher as closer the
distance between nodes in a link, but this common sense does
not properly guarantee the actual quality of network servicedue to interference caused by other competing links. Fig. 2
shows the configuration of a simple experiment that presents
the phenomenon. In the experiment, an AP and a laptop is
composed as one communicating pair, and distance between
14
8/10/2019 QoE-driven Wi-Fi Selection Mechanism for Next Generation Smartphones
3/6
3
Fig. 2: Channel interference problem
them is about a meter without any obstacle between them.One pair surrounded by solid line box is configured to use
the channel 1 and the other two pairs surrounded by dashed
line box is set to use the channel 6 in common. No other linkexists in the proximity of this experimentation and only one
smartphone eavesdrops signal to measure RSSI for each beacon
of AP. Each laptop communicates with the server by sending
iperfmessage for 1000 seconds using the established link with
the closest AP. The laptop using the channel 1 alone showed
throughput of 21 Mb/s, whereas the other two pairs showed13 to 14 Mb/s individually. This is because the channel 6 isused by two laptops within the same hop and therefore each
station interferes with each other. The pair with the channel 1
is not interfered because the channel 1 and 6 are completelynon-overlapping channels. However, RSSI values for all APs
beacon messages were about -40 to -45dBm in perspective ofthe eavesdropping smartphone device. This experiment clearlyshows that RSSI is a not good indicator to be used as a metric
for wireless Internet service quality. And it is trivial to expect
that Wi-Fi connection that neglects quality metric is expected
to result in even worse Internet services to device users.
Above three problems show that there are three importantaspects that should be addressed for optimal AP selection:
A full scanning is required; device should collect all beaconmessages by fully scanning all channels in order to select
the best AP within the sensing range of the device;
Authentication-free APs are candidates in selecting suit-able APs; more AP choices bring higher probability of
connecting the Internet and selecting qualified AP;
Medium access control (MAC) protocol should be used as
a selection criterion regardless of whether it is contentionbased or contention-free; the number of stations using
the same channel should be counted to see how many
communications are made with the same channel.
Considering above aspects, we propose the optimized APsearch and selection mechanism, and we also validate the per-formance improvement that the proposed mechanism achieves.
III. QOE- DRIVENA P SELECTION M ECHANISM
The proposed AP selection mechanism for maximizing usersquality of experience (QoE) consists of two important schemes.
The first scheme is the ISP & Authentication based AP can-
didate selection scheme that is used to increase the probability
of connecting a qualified AP. The second scheme is the SINR-
Fig. 3: A simulation topology for ISP & Authentication search
based AP selection scheme that is used to connect the optimal
AP among connectable candidate APs.
A. ISP & Authentication based AP Candidate Selection
Since smartphones usually connect to an AP that is already
configured in its small database, some service providers inscribeits configuration data into the phone database before customers
use it, and thus, APs of contracted service provider are automat-
ically connected by the device without any notification to users.
However, authentication-free APs cannot be automatically usedby the device, and therefore AP candidate selection should take
into account those APs as candidates for the connection.
We conducted a simple simulation to show the impact of
having more choices for connection. The network configuration
consists of 12 APs and one hand-held device in the way that
12 APs surrounds the hand-held device. As for the device, itreceives data from a server through one of the APs. As for
APs, one fourth of those APs belong to a contracted serviceprovider, another one fourth of APs are free to be used. The
rest of APs are unusable because they are secured with some
authentication mechanism or belong to other service providers
The server sends 1 Mb/s downlink IP flow to the device andthe connected AP is supposed to relay the flow to the device
Fig. 3 describes the simulation setting visually.
In the simulation, we compared two results where the one
comes from the original connection mechanism built in the
OPNET simulator and the other is obtained when we useauthentication-free APs. In order to see the pure effect of the
authentication-free AP considered mechanism, we do not useany scheme that qualifies the link status of each AP. Recalthat as mentioned early, the first arrived beacon message that
contains configured AP information triggers the connection
establishment process in the original mechanism.
We employed MAC layer throughput as a performance metric
to see the impact of including free APs as candidates for thenetwork connectivity. When we used authentication-free APs for
the connectivity, we observed that 51.9 % of MAC throughput
has increased in average compared to the original connection
mechanism.
15
8/10/2019 QoE-driven Wi-Fi Selection Mechanism for Next Generation Smartphones
4/6
4
It is general that most of APs require an authentication
process to network users for access authorization but cafesand restaurants around public places do provide the Internet
service to customers by opening APs without any authentication
requirements. This simulation results show that Internet service
providers are not only those telecom companies but also cafe& restaurant owners that provide Internet access while selling
their products. The reason of providing APs is because many
customers now use hand-held devices to access the Internet.Therefore, neglecting authentication-free APs as a candidate is
not a good idea in those places.
B. SINR based optimal AP Selection scheme
After gathering candidates via the above scheme, the opti-
mized AP selection scheme is conducted to connect the best
qualified AP. The original connection mechanism does not useany quality metric to select an AP, and some of the Wi-Fi related
applications refer to RSSI value and suggest the user to connect
an AP with the highest RSSI value. However, as mentionedearlier, RSSI-based network selection sometimes results in bad
QoE. For example, even if an AP with the channel 1 shows
the largest RSSI of all, meaning the AP is positioned nearby auser, the channel 1 can be possibly used by many other APs
and wireless devices that communicate to those APs. Based on
this observation, we proposes SINR based optimal AP selection
scheme that considers the number of stations communicating ateach channel in order to make an wireless device connect into
the least crowded channel. We use the following equation to
measure the quality of each AP or estimate SINR for each AP:
SINR =RSSIcur
xSLxRSSIx +N
Note that RSSI value in the equation is initially measured indBm but it is translated intomiliwatts[7]. Nominator in equa-tion III-B shows the RSSI value of AP to be currently evaluated,
whereas denominator indicates the summation of RSSIs fromAPs that use the same channel. x indicates an AP within the
set of S and S is the set of APs that use the same channel. L
indicates the number of stations that communicate through theAP x. N indicates the noise level measured while receiving a
beacon frame from the current AP. Information of the channel
number and the number of communicating stations is containedin the beacon frame received from each AP, and the RSSI value
can be measured while receiving beacon physically. Note that
the number of stations connected to each AP is described in the
22nd element of the beacon body named BSS Load. The BSSLoad value contains the number of stations using corresponding
AP and the channel utilization value provided as percentage of
how many busy time slot is counted among all time slots. TheBSS load is frequently used in articles that focused on load
balancing among nearby APs [8], [9]. The channel information
can be earned by PHY parameter sets or catching the momentof receiving beacon and see what channel is scanned. Also, the
channel number of each AP can be easily earned by Android or
iOS SDK using scan or Wi-Fi related libraries. The noise level isalso available at devices antenna, and some user level programs
like wavemon support the functionality of showing noise level
of specified AP in real time. The available AP set Salso includes
the current AP, so that nominator of equation indicates the
signal power of the link which can be possibly generated by
the downlink flow from the AP to the device that attempts toconnect. This algorithm is designed in the assumption that most
data flows stream from AP to devices. It means that the RSSI
value of AP can be represented as the power generated by that
AP because APs are the ones that take over channel the mostin case of downlink flow. In this paper, SINR is defined as the
smallest possible SINR that can happen to the users device if
it is connected to the AP. The SINR based selection providesa new priority based metric that can reinforce the RSSI metric
to provide an optimal AP for users according to the degree
of crowd of each channel. However there are several issuesrevealed as follows:
Interference caused by other stations can be diminished ifdevices that cause uplink flows are out of carrier sensing
range; If all devices and APs are within one hop range, the
SINR metric is applicable to both downlink and uplink butthis rarely happens in reality;
WLAN channels are partially overlapped with each other
so that the number of completely non-overlapping channelis only three, compared to total 13 channels in 2.4GHz
band; Adjacent-channel interference [10] can cause smallerSINR value estimated by the proposed equation;
AP with very low RSSI that has no competitors in the samechannel can possibly have larger SINR compared to other
APs with much larger RSSI and some competitors.
Uplink flows are lightly considered in this paper since most
of the Internet service data flow from servers to users. Adjacent-channel interference can be solved by AP installers in the way
that service providers can install APs that possess completely
non-overlapping channel to minimize interference. AP with very
low RSSI problem can be solved by giving a threshold of RSSIvalue to filter out useless APs. IEEE 802.11 protocol release at
the year of 2007 specifies receiver performance requirements
and these requirements give the minimum RSSI value for usageof each modulation technique. RSSI value less than -82 dBm
can not be used for any modulations used by 2.4GHz Wi-Fi in
current technology. Threshold values of -70 to -82 dBm resultedin good performance of using SINR metric. Simulation results
described in this paper used -75dBm as lower bound threshold
of using as a connection candidate.
IV. PERFORMANCEE VALUATION
In this section, we implemented the proposed qualified APselection mechanism in both a Android phone and the OPNET
simulation. Then we carried out a performance evaluation study
to verify the effectiveness of the proposed scheme.
A. Simulation Study
The simulation configuration and scenarios are configured
as follows to verify if the proposed qualified AP selectionmechanism improves the network throughput as follows:
We placed in the simulation network twelve pairs of AP &
STA using channel 1, 6 or 11; channel number is uniformly
distributed to APs and STAs;
We applied the proposed mechanism to the 13-th STA
and there is no pair AP to this STA at the beginning of
simulation; therefore,this STA has 12 connection choices
at that time;
16
8/10/2019 QoE-driven Wi-Fi Selection Mechanism for Next Generation Smartphones
5/6
5
(a) The original mechanism (b) The proposed mechanism
Fig. 4: MAC throughput comparison between the original and proposed mechanism
Fig. 5: Simplified simulation topology
We configured 6 out of 12 APs to be connectable to the
STA, which are either APs of contracted Internet service
provider or authentication-free APs; rest of APs are notcandidates for connection;
We conducted 25 simulation runs with different combina-
tions of AP & STA positions and channel assignments:
We used 5 different positioning of AP & STA location;
We used 5 different channel assignment to each pair.
Fig. 5 presents a simplified simulation environment. Note thatthe real deployment in the OPNET simulation is not as regular
as shown in the figure. As shown in the figure, each device
are not guaranteed to be connected to the nearest AP and eachAP can contain more than one wireless device or none. Noise
level is constant through all channels since there is no jammer
installed in the environment. Fig. 4 shows the comparison ofperformance between two mechanisms: the original mechanism
and the proposed mechanism. Original mechanism produces bad
MAC throughput generally since the number of connectablecandidates is small and no qualification metric is used for the
connection. On the other hand, the proposed QoE driven Wi-
Fi selection mechanism produces much greater performance
in some cases. We have the following observations from the
simulation results: (i) The proposed mechanism improved MAC
throughput by 255% in average. Even if extraordinary perfor-
mance improvement in the channel assignment 5 is not included,
the MAC throughput is still improved by 89.9% in averageThe reason why the channel assignment 5 makes the best
improvement is because the original mechanism connected the13th device to an AP that has most number of stations, while
the proposed mechanism connected the device to an AP with
the least number of stations. (ii) There were cases that shows
no difference in usage between the original or the proposedmechanism because the original mechanism sometimes connects
to the optimal AP (that the proposed mechanism decides)
accidentally. Also, (iii) the performance degradation happenedat some cases. This is not because the mechanism failed to
select the optimal AP but because IEEE 802.11 MAC protocol
makes serious starvation in a particular receiver. The issue is
out of scope of this paper.
B. Android Application
The proposed mechanism has been implemented in AndroidOS based Smartphones to determine AP candidates with esti-
mated SINR values. Android SDK provides the result of Wi-
Fi scanning and each scanned AP contains SSID information,capabilities and RSSI values. However, BSS Load information
is not shown since the SDK can only distribute a small part of
beacon frame. Therefore, the current implementation uses rather
the number of APs in the same channel instead of the numberof stations in the channel. Fig. 6 shows the process of using the
application ((a), (b)) and a map that shows location spots where
an experiment was conducted ((c)). Four buttons are distributedon the screen. User clicks the SSID button to setup SSID for ISP
identification and clicks the start button to begin AP scanning.
Screen continuously updates the result of scanning but user canclick the stop button to stop the scanning. Screen presents the
number of scanned APs, the number of APs that exceed signal
level of -80 dBm, the number of connectable AP candidates,and list of candidate APs with information of channel, SINR and
RSSI. Application suggests the user to connect an AP with the
highest SINR value and shows what the order of APs prioritized
by SINR is. User can manually select an AP to connect, or the
17
8/10/2019 QoE-driven Wi-Fi Selection Mechanism for Next Generation Smartphones
6/6
6
(a) SSID input interface (b) Scanning result view (c) Scanned locations
Fig. 6: Android experiment
TABLE I: Android experiment results
Location Index Nscanned Nsignal Ncandidates
1 12 6 22 22 10 33 21 11 44 10 5 15 12 3 36 8 2 2
application can automatically connect to the AP with the highestSINR because the configuration history of APs of contracted
service provider has been already saved to the phone, and free
APs can be connected without any additional requirements.
Fig. 6 (c) shows 6 Locations which are evaluated in terms
of the number of AP candidates. Application first filters outAPs with too low beacon signals, and it makes a candidate list
of APs of contracted service provider and authentication-free
APs. Table I shows the results of this experiment. Note thatlocation index in the table starts from the bottom of the map
(Fig. 6 (c)). We can observe that a few APs are connectable
even though there are a lot of APs installed around device user.
V. CONCLUSION
We proposed a QoE driven Wi-Fi selection mechanism in this
paper. The proposed mechanism is conducted with two stages:ISP & Authentication based AP candidate selection is carried
out at the first stage to increase the availability of network
connectivity by considering authentication-free APs, and thenSINR based AP selection is performed at the second stage to
choose the best qualified AP. To validate the effectiveness of the
proposed mechanism, we realized it in a Android Smartphoneand also implemented a OPNET simulation environment, and
then we conducted a performance evaluation study. The perfor-
mance results indicate that the proposed mechanism can supportan optimal Wi-Fi connection and also provide better Internet
service for wireless users. Based on the results, we vision that
the proposed mechanism can be equipped at the next generation
smartphones and it can aid vertical handover among various
networks. We have several research plans as future work. We
will relax the downlink assumption for the SINR estimationWe also would like to carry out an extensive empirical study to
verify if the proposed mechanism can handle various networktopologies and traffic patterns.
ACKNOWLEDGMENT
This research was supported in part by the KCC (Korea Communications
Commission), Korea, under the R&D program supervised by the KCA (Korea
Communications Agency) (KCA-2012-08-911-05-001), and in part by the
National Research Foundation of Korea (NRF) Grant funded by the Korea
government (MEST) (No. 2010-0014060).
REFERENCES
[1] J. Mo and W. Kim, Choosing internet pricing schemes: Flat or usage inkorea, International Telecommunications Policy Review, Vol. 18, No. 12011, 2011.
[2] J. Reynolds, Going Wi-Fi: a practical guide to planning and building an802.11 network. CMP, 2003.
[3] IEEE Standard for Local and metropolitan area networksPart 21: mediaIndependent Handover Services, IEEE Std.
[4] S. Rizvi, A. Aziz, N. Saad, N. Armi, and M. Yusoff, Tight couplinginternetworking between umts and wlan: Challenges, design architecturesand simulation analysis, International Journal of Computer Networks(IJCN), vol. 3, no. 2, p. 116, 2011.
[5] IEEE Standard for Informantion technologyTelecommunications andinformation exchange between systemsLocal and metropolitan areanetworksSpecific requirementsPart 11: Wireless LAN Medium AccessControl (MAC) and Physical Layer (PHY) Specifications, IEEE Std.
[6] H. Cho, S. Bae, H. Choo, and M. Chung, Implementation of wlan connection management schemes in heterogeneous network environments withwlan and cellular networks, Computational Science and Its Applications-
ICCSA 2011, pp. 3243, 2011.[7] D. Green and A. Obaidat, An accurate line of sight propagation per
formance model for ad-hoc 802.11 wireless lan (wlan) devices, inCommunications, 2002. ICC 2002. IEEE International Conference onvol. 5. IEEE, 2002, pp. 34243428.
[8] E. Garcia, J. Ferrer, E. Lopez-Aguilera, R. Vidal, and J. Paradells, Clientdriven load balancing through association control in ieee 802.11 wlans,Transactions on Emerging Telecommunications Technologies, vol. 20no. 5, pp. 494507, 2009.
[9] E. Garcia, R. Vidal, and J. Paradells, Cooperative load balancing in ieee802.11 networks with cell breathing, in Computers and Communications2008. ISCC 2008. IEEE Symposium on. Ieee, 2008, pp. 11331140.
[10] V. Angelakis, S. Papadakis, V. Siris, and A. Traganitis, Adjacent channeinterference in 802.11 a is harmful: Testbed validation of a simplequantification model, Communications Magazine, IEEE, vol. 49, no. 3pp. 160166, 2011.
18