See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/348168218

A new SDN-based load balancing algorithm for IoT devices

Article in Indonesian Journal of Electrical Engineering and Computer Science · February 2021
DOI: 10.11591/ijeecs.v21.i2.pp1209-1217

CITATION READS

1 128

3 authors:

Hind Sounni Najib EL KAMOUN

Université Chouaib Doukkali Université Chouaib Doukkali
9 PUBLICATIONS 15 CITATIONS 189 PUBLICATIONS 548 CITATIONS

SEE PROFILE SEE PROFILE

Fatima Lakrami
Université Chouaib Doukkali
30 PUBLICATIONS 64 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

SD-Access ( Fabric Network, Automation and Analytics LAN ) - Campus Networks View project

Novel Frameworks of Smart and Adaptive QOS Management for computer networks View project

All content following this page was uploaded by Hind Sounni on 03 January 2021.

The user has requested enhancement of the downloaded file.

Indonesian Journal of Electrical Engineering and Computer Science
Vol. 21, No. 2, February 2021, pp. 1209~1217
ISSN: 2502-4752, DOI: 10.11591/ijeecs.v21.i2.pp1209-1217  1209

A new SDN-based load balancing algorithm for IoT devices

Hind Sounni, Najib El kamoun, Fatima Lakrami

STIC laboratory, Science Faculty, Chouaib Doukkali University (UCD), Morroco

Article Info ABSTRACT

Article history: Nowadays, the emergence of the Internet of things devices has wholly
revolutionized the customer's communication habits. The information can be
Received Jun 15, 2020 collected at any time and anywhere. However, the mobility of
Revised Aug 30, 2020 communication devices in a dense network results in several problems, such
Accepted Sep 27, 2020 as unbalanced network load and increased bandwidth demands, which
decrease the network performance. To deal with these issues, this paper
proposes a new load balancing algorithm based on the software-defined
Keywords: Network to enhance the performance of mobile devices communication over
a Wi-Fi network. The use of the software-defined network automatizes the
Internet of things (IoT) configuration of the network through a centralized controller; it provides
Load balancing programmability and an overall view of the network, along with optimizing
Mobility management resource allocation based on real-time network information, which facilitates
Software-defined network the implementation of our algorithm. The proposed algorithm is implemented
Wi-Fi and evaluated through simulation using mininet-WiFi. The results indicate
that our proposed method provides an efficient network load balancing and
improves performancdevices' performance.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Hind Sounni
STIC Laboratory
Chouaib Doukkali University
El Jadida, Morocco
Email: sounni.h@ucd.ac.ma

1. INTRODUCTION
Internet of Things (IoT) [1], or in a broader sense, Internet of Everything (IoE), is a relatively recent
concept. It is considered a significant technological and economic innovation in new information and
communication technology. IoT is defined as an expansion of today's internet to all devices that can directly
or indirectly communicate with electronic devices that are connected to the internet. Many heterogeneous
devices (telephone, computer, watch, refrigerator) are now widely used in everyday life. Figure 1 shows that
the most commonly used IoT architecture comprises three layers: the perception layer, the network layer, and
the application layer.
With the exponential development of these connected devices with heterogeneous characteristics,
future networks must evolve towards new architectures to adapt to the increase in traffic load, ensure
security, and improve service quality, especially in mobile devices. The standardized communication
protocols used for IoT, such as 802.11 standards [2], do not offer native mobility management or load
balancing mechanisms. In a traditional Wi-Fi network, the re-association process between APs and nodes is
based on the RSSI (Received Signal Strength Indication) [3]. This can lead to an unbalanced load of the
different access points, especially when moving within the network, the mobile node may connect to an
overloaded access point. Besides, the authentication process between the node and the AP during the re-
association is a time-limited phase. If this delay is exceeded, a loss of network connection occurs. These
limitations may result in degradation and reduction of applications running on IoT device's performances.

Journal homepage: http://ijeecs.iaescore.com

1210  ISSN: 2502-4752

In this context, developing efficient solutions for load balancing is an excellent necessity for better
performance. In [4], the authors propose a new algorithm based on the subgradient method, called
Congestion-Optimal Wi-Fi Offloading (COWO), to achieve the ideal offloading ratio for the different access
point to ensure maximized throughput and minimizes the traffic congestion. In [5], a new architecture is
presented for a seamless handover and a better load balancing using SDN. The authors provide a solution to
minimize data propagation delay through the earlier publication of the OpenFlow table. Using a theoretical
model, the authors in [6] introduce a new inter-BSS handoff scheme for the Wi-Fi network where the mobile
nodes switch from one AP to another in an overlapping area, especially when detecting a performance
degradation. An interleaved scanning mechanism is conceived to identify the alternative APs in the proximity
of a mobile node. Authors in [7] propose a cell breathing technique for load balancing in WLAN, especially
the Wi-Fi network, to prevent network traffic congestion and APs saturation. Two algorithms are used to
minimize the load of the loaded AP(s) in the network, and the second generates an optimal load-balanced
solution. In [8], LABERIO is presented to solve the load imbalance issue caused by the loading failure in the
middle of traffics’ transmission. To evaluate the packet transmission path and adjust the ways to achieve load
balance dynamically, authors in [9] use the Fuzzy Synthetic Evaluation Mechanism. Authors in [10] propose
to distribute the load by implementing an algorithm based on a round-robin. The simulation results show that
the proposed method's response time gives better results than the random load balance. The authors in [11]
implement a load targeted handover (LTH) scheme to enhance mobility management by offering a seamless
handover and balance the load by a targeted association with less loaded APs. In [12], the authors used a
method based on the server response time; the server with the most stable response time is selected to ensure
network load balancing. Authors in [13] propose a two-tier dynamic load balancing solution for the SDN-
enabled Wi-Fi network is presented; this method allowed the controller to estimate the load balancing degree
among the different APs and determine until which load level the APs can accept association requests
without consulting the controller. In comparison, in our paper, we can ensure good mobility management
using the SDN controller and balance the network load using our load balancing algorithm.
In recent years, the introduction of Software-Defined Networks (SDNs) has introduced new technics
for enhancing network and communications management [14]. SDN is considered a significant innovation
for more efficient management of connected devices and can meet the IoT's current needs for heterogeneity
and flexibility. The SDN enables the separation between the control plane and the data plane and provides
centralized control to facilitate the optimization and configuration efficiently and automated. It also
accelerates the deployment of new services by responding dynamically to policy changes [15]. The use of
SDN as a centralized architecture enforces SLAs by establishing a unified network that allows users to have
an equitable experience regardless of their access methods; it also ensures interoperability between
heterogeneous devices, which make it very suitable for IoT networks. All the cited enhancements enable us
to provide better results by improving network quality for the IoT.

Figure 1. IoT architecture

In this paper, we propose a load balancing algorithm based on SDN for enhancing the performance
of mobile IoT devices communication over a Wi-Fi network. Our objective is to highlight the contribution of
deploying SDN for managing the high traffic load generated in an IoT architecture and improving network

Indonesian J Elec Eng & Comp Sci, Vol. 21, No. 2, February 2021 : 1209 - 1217
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752  1211

performances by enhancing the throughput of the different APs, reducing the packet loss and the
communication delay. We propose an extensive simulation using the Mininet Wi-Fi emulator to evaluate and
compare the proposed approach's performance with the traditional network architecture using standard AP
(Access Point). The rest of the paper is structured as follows. Section 2 introduces the proposed approach.
Section 3 describes the simulation setup. Section 4 provides the simulation results and discussion, while
Section 5 concludes the paper and offers suggestions on future researches.

2. PROPOSED ALGORITHM
Several technologies can be deployed for connecting IoT devices at the physical level [2, 16-19].
The new 802.11ax standard [20] is considered the most suitable standard for IoT, allowing high throughput
and reducing the connected devices' energy consumption. One of the significant challenges in implementing
the Wi-Fi network is mobility. Most mobile devices need to be connected to the services they want at all
times while moving; the interruption of these services can occur when they move from one access point to
another. In a traditional Wi-Fi network, the association process between APs and nodes is based on the RSSI;
this can lead to an unbalanced load of the different access points. The mobile node may connect to an
overloaded access point, especially when moving within the network. This process has a significant impact
on the performance of the entire Wi-Fi network.
In this article, we integrate SDN technology to improve both mobility management and network
load. Through the separation of the control plane from the data plane, the SDN makes the network's
automatic configuration via the controller and allows efficient management of scalability and mobility of the
equipment, as we demonstrated in our previous work [21]. The centralized controller also provides
programmability, a global view of the network, and optimizes resource allocation based on real-time network
information that helps implement rules and algorithms to optimize Wi-Fi network performance.
In the proposed algorithm, the controller collects information such as AP capacity, current load of
APs, the distance separating different APs, and the lists of devices associated with each AP. This information
helps the controller to calculate and deduce the best AP for the new associations. When a device enters an
overlapped area, it broadcasts probe request messages; the targeted AP collects the connected devices'
information within this message and sends it to the controller. The controller can refuse or accept the
association's request of the connected devices to the specific AP based on the collected information. The
controller will redirect the device to another AP if the targeted AP is overloaded. The OpenFlow switches
reside in the access points; they are responsible for the mobility management of the tools and the exchange of
OpenFlow messages with the controller via the OpenFlow protocol [22].
As shown in Figure 2, when a mobile device is entering an overlapping area, it will be associated
with an AP based on the RSSI. The controller will compare the load of this AP to a threshold. If it detects
that this access point is overloaded, it checks the possibility of transferring the connected device to a
neighboring AP (the less loaded AP). In this regard, we are working on the same process evoked by [23] by
adding a significant step, adjusting each AP's transmission power according to their load. The controller
sends a message (beacon frame) to the first AP to disconnect the device and adjusts the power of the other
APs so that the less loaded AP's strength is higher than the power of other APs in the overlapping area. The
disconnected device does not need a handshake message to connect to a new AP because its MAC address is
already stored in the controller database. For no-mobile devices located in an overlapped area, the AP's
association is also based on the strongest RSSI. The controller will accept the association requests if this AP
is less loaded. Otherwise, the device will be redirected to another AP. This process is both fast and
straightforward and will improve the network load and its throughput.

Figure 2. Description of the SDN-based load balancing algorithm

A new SDN-based load balancing algorithm for IoT devices (Hind Sounni)
1212  ISSN: 2502-4752

Figure 3 describes the association control message flow between mobile devices, OpenFlow
enabled-APs, and the SDN controller. The APs are connected to the SDN controller via a secure channel. We
suppose that the mobile device is located in an overlapping area covered by the three access points (AP1,
AP2, AP3), and based on RSSI, it is connected to one of the three APs (for example, AP1).
a) Step 1. When a device enters the overlapping area, it receives a beacon frame from the three APs and
broadcasts probe request messages.
b) Step 2. The APs reply with the probe response messages and report via the Packet-In message to the
controller: their load value, association event, also the device information contained in the probe
request.
c) Step 3. The controller updates the APs' information periodically, checks the AP load and association
events and runs the load balancing algorithm if necessary (AP load>threshold).
d) Step 4. If the (AP1 load > threshold), the controller sends beacon-config messages (SetConfig message)
to the AP1 to disassociate the device and adjusts the power of the less loaded AP in such a way to be
higher than the other APs in the overlapping area.
e) Step 5. The AP1 sends a disassociation notification to the device.
f) Step 6. The same as step 1.
g) Step 7. The device sends an association request to the AP with the high RSSI; once the device receives
the AP's association response, the connection is established.
Setting the signal power values of the different APs according to their load (step 4) is crucial since it
improves the association time between devices and access points, especially in mobile devices located in an
area covered by more than 2 APs. Our algorithm is developed to allow the controller to configure the least
loaded access points with the highest signal strength. In this case, the device will directly re-associates with
the appropriate AP, and it will not waste time trying to connect to an overloaded access point and switch to
another. This will optimize the association time, and thus the number of requests exchanged.

Figure 3. Message Flow for association using load balancing algorithm based SDN

3. SIMULATION SETUP
The SDN architecture is a distributed design in which a centralized global controller is used to
monitor other equipment’s within the network. The application resides on the top of the SDN controller, as

Indonesian J Elec Eng & Comp Sci, Vol. 21, No. 2, February 2021 : 1209 - 1217
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752  1213

illustrated in Figure 4, and it uses different interfaces to implement network services. Mobile stations are
accessing services through OpenFlow-enabled APS, which are controlled by the SDN controller.

Figure 4. SDN-based

To simulate the wireless network scenario with SDN and to demonstrate the performance of our
proposed method, we used: mininet Wi-Fi version 2.2.2 [24] running on Ubuntu 16.04, VMware software
tool to run the virtual network, and Iperf [25] to generate TCP, and UDP traffic flows between the different
devices [26]. The detailed simulation parameters are given in the Table 1:

Table 1. Simulation setup setting

Controller RYU v4.27
Operating system Ubuntu 16.04
Type of traffic TCP, UDP
Packet size range Min 1 Mb
Max 50 Mb
APs 802.11 ax {AP1, AP2, AP3}
Tx power {20 dbm, 20 dbm, 20 dbm}
SSID {new-ssid1, new-ssid2, new-ssid3}
Mobility speed 1 m/s

As shown in the Figure 5, the three APs are set so that their coverage overlaps. Mobile and fixed
stations represent the smart devices; each station is given a unique IP and Mac address. The whole topology
is under the control of the RYU, an open-source controller, licensed by Apache 2.0, entirely written in
Python, supported and deployed by Nippon Telegraph and Telephone's cloud data centers [27].

A new SDN-based load balancing algorithm for IoT devices (Hind Sounni)
1214  ISSN: 2502-4752

Figure 5. Simulated scenario

4. SIMULATION RESULTS AND DISCUSSION

Our first scenario includes 30 stations associated with three access points based on the RSSI value
(traditional method), six stations are associated with AP1, 14 with AP2, and 10 with AP3. These stations
exchange TCP traffic. Figure 6(a) presents the result of the first scenario and shows that the network load is
unbalanced: the station connected to AP2 receives a lower throughput (51 Mbps on average) compared with
AP1 (56 Mbps) and AP3 (63 Mbps). In the second scenario, we use the same topology, but with the
implementation of our load balancing algorithm based on SDN, the results show that the network load is
balanced and that the number of stations is equally distributed among the three APs, which improves the
throughput in each AP as shown in Figure 6(b).

(a) (b)

Figure 6. TCP traffic in traditional Wi-Fi network and when using the
load balancing algorithm based on SDN

To better evaluate the proposed approach, we decided to exchange UDP traffic and perform the
same experiment with the same parameters. As shown in Figure 7(a), the stations connected to AP2 receives
a low throughput compared with AP1 and AP3. After deploying our load balancing algorithm, the load is
balanced between the three APs as shown in Figure 7(b).

Indonesian J Elec Eng & Comp Sci, Vol. 21, No. 2, February 2021 : 1209 - 1217
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752  1215

(a) (b)

Figure 7. UDP traffic in traditional Wi-Fi network and when using the
load balancing algorithm based on SDN

Figure 8(a) presents the jitters variation in traditional and using our load balancing algorithm based
on SDN. In a traditional network, the average jitter value during the UDP transmission for stations associated
with AP1 is 0,64 ms, the stations related to AP2 and AP3 receive 2,14ms and 1,62 ms, respectively. When
using our load balancing algorithm based on SDN, the jitter values of the three APs (AP1, AP2, and AP3) are
1,50 ms, 1,59 ms, and 1,54 ms. An enhancement of the jitter values is observed in Figure 8(b) in comparison
to the traditional network.

(a) (b)

Figure 8. Jitter values in traditional Wi-Fi network and when using the
load balancing algorithm based on SDN

5. CONCLUSION
In this paper, we proposed a load balancing algorithm based on SDN for enhancing the performance
of mobile IoT devices communication over Wi-Fi networks. We built a simulation environment using
mininet to implement and analyze the performance of the developed concept. The results have shown that the
proposed scheme significantly balances the entire network's load and enhances the throughput of the devices
compared with the traditional architecture where no load balancing mechanism is applied. We aim to
consider other constraints and metrics in developing and evaluating the current algorithm in the future.

REFERENCES
[1] J. H. Nord, A. Koohang, et J. Paliszkiewicz, “The Internet of Things: Review and theoretical framework,” Expert
Systems with Applications, vol. 133, p. 97-108, 2019.
[2] Q. Qu et al., “Survey and Performance Evaluation of the Upcoming Next Generation WLAN Standard - IEEE
802.11ax”, Mobile Networks and Applications, p. 155, 2019.

A new SDN-based load balancing algorithm for IoT devices (Hind Sounni)
1216  ISSN: 2502-4752

[3] K.-L. Yap, Y.-W. Chong, et W. Liu, “Enhanced handover mechanism using mobility prediction in wireless
networks”, PLOS ONE, vol. 15, no 1, p. e0227982, 2020.
[4] L.Bin, Z.Qi, T.Weiqiang, and Z.Hongbo.“Congestion-Optimal Wi-Fi Offloading with User Mobility Management
in Smart Communications”. Hindawi Wireless Communications and Mobile Computing, Article ID 9297536, no.
15, 2018.
[5] S. M. M. Gilani, T. Hong, W. Jin, G. Zhao, H. M. Heang, and C. Xu, “Mobility management in IEEE 802.11
WLAN using SDN/NFV technologies”, Eurasip Journal on Wireless Communications and Networking, vol. 2017,
no. 1, 2017.
[6] Sarma, S. Chakraborty, S. Nandi, et A. Choubey, “Context Aware Inter-BSS Handoff in IEEE 802.11 Networks:
Efficient Resource Utilization and Performance Improvement”, Wireless Pers Commun, vol. 77, no. 4, pp.
2587‑2614, doi: 10.1007/s11277-014-1656-8, 2014.
[7] S. K. Singh, M. Tech, et M. Madhavi, “Cell Breathing Technology in Wi-Fi/Lans”, International Journal of
Computer Trends and Technology, vol. 4, no. 9, p. 6, 2013.
[8] H. Long, Y. Shen, M. Guo, F. Tang “LABERIO: dynamic load-balanced routing in OpenFlow-enabled networks”,
IEEE 27th International Conference on Advanced Information Networking and Applications (AINA), pp. 290-297,
2013.
[9] J. Li, X. Chang, Y. Ren, Z. Zhang, G. Wang “An effective path load balancing mechanism based on SDN”, 2014
IEEE 13th International Conference on Trust, Security and Privacy in Computing and Communications, pp. 527-
533, IEEE 2014.
[10] S. Kaur, K. Kumar, J. Singh, N.S. Ghumman “Round-robin based load balancing in software defined networking”
2nd International Conference on Computing for Sustainable Global Development (INDIACom), pp. 2136-2139,
2015.
[11] S. M. M. Gilani, W. Jin, T. Hong, G. Zhao, et C. Xu, “SDN-based Handover in Future WLAN”, IJFGCN, vol. 9,
no 12, p. 139 154, doi: 10.14257/ijfgcn.2016.9.12.13, 2016.
[12] H. Zhong, Y. Fang, J. Cui “LBBSRT: an efficient SDN load balancing scheme based on server response time”,
Future Generat. Comput. Syst., 68, pp. 183-190, 2017.
[13] Y.-D. Lin, C. C. Wang, Y.-J. Lu, Y.-C. Lai, et H.-C. Yang, “Two-tier dynamic load balancing in SDN-enabled Wi-
Fi networks”, Wireless Netw, vol. 24, no 8, p. 2811 2823, doi: 10.1007/s11276-017-1504-3, 2018.
[14] Z. Chen, Z. Luo, X. Duan, et L. Zhang, “Terminal handover in software-defined WLANs”, J Wireless Com
Network, vol. 2020, no. 1, p. 68, 2020.
[15] W. B. Jaballah, M. Conti, et C. Lal, “A Survey on Software-Defined VANETs: Benefits, Challenges, and Future
Directions”, arXiv:1904.04577 [cs], 2019.
[16] U. N. Kar et D. K. Sanyal, “An overview of device-to-device communication in cellular networks”, ICT express,
vol. 4, no. 4, pp. 203-208, 2018.
[17] S. Zeadally, F. Siddiqui, et Z. Baig, “25 years of bluetooth technology”, Future Internet, vol. 11, no 9, p. 194, 2019.
[18] O. P. Bodunde, U. C. Adie, O. M. Ikumapayi, J. O. Akinyoola, et A. A. Aderoba, “Architectural design and
performance evaluation of a ZigBee technology based adaptive sprinkler irrigation robot”, Computers and
Electronics in Agriculture, vol. 160, p. 168-178, 2019.
[19] R. Jain, “Low Power WAN Protocols for IoT: IEEE 802.11 ah, LoRaWAN, Sigfox” Lectures,Washington
University, 2018.
[20] D. López-Pérez, A. Garcia-Rodriguez, L. Galati-Giordano, M. Kasslin, et K. Doppler, “IEEE 802.11 be extremely
high throughput: The next generation of Wi-Fi technology beyond 802.11 ax”, IEEE Communications Magazine,
vol. 57, no. 9, p. 113-119, 2019.
[21] S. Hind; E.Najib, L. Fatima, “Towards Qos enhancement in wireless network through SDN”, in 2019 7th
Mediterranean Congress of Telecommunications (CMT), p. 1-5, 2019.
[22] S. Hind; E.Najib, L. Fatima, “Software Defined Network for Qos Enhancement in Mobile Wi-Fi Network”, IJRTE,
vol. 8, no 3, p. 4863 4868, doi: 10.35940/ijrte.C6889.098319, 2019.
[23] N. Kiran, Y. Changchuan, et Z. Akram, « AP load balance based handover in software defined Wi-Fi systems », in
2016 IEEE International Conference on Network Infrastructure and Digital Content (IC-NIDC), pp. 6‑11, 2016.
[24] R. R. Fontes, S. Afzal, S. H. Brito, M. A. Santos, et C. E. Rothenberg, “Mininet-WiFi: Emulating software-defined
wireless networks”, in 2015 11th International Conference on Network and Service Management (CNSM), pp. 384-
389, 2015.
[25] J. Dugan, “Iperf Tutorial “, Columbus: Summer JointTechs, pp. 1-4, 2010.
[26] M. T. Naing, T. T. Khaing, et A. H. Maw, “Evaluation of TCP and UDP Traffic over Software-Defined
Networking”, in 2019 International Conference on Advanced Information Technologies (ICAIT), pp. 7-12, 2019.
[27] S. Asadollahi, B. Goswami, et M. Sameer, « Ryu controller’s scalability experiment on software defined
networks », in 2018 IEEE International Conference on Current Trends in Advanced Computing (ICCTAC),
Bangalore, pp. 1‑5, 2018.

Indonesian J Elec Eng & Comp Sci, Vol. 21, No. 2, February 2021 : 1209 - 1217
 Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752  1217

BIOGRAPHIES OF AUTHORS

Hind Sounni received the Master degree in Networking and Telecommunications, from science
faculty, at Chouaib Doukkali University, El Jadida, Morocco in 2016. Actually, she is a Ph. D
Student in Networks and Telecommunication at STIC Laboratory in the Science Faculty at
Chouaib Doukkali University. Her research interests include: Wireless network, Qos in WLAN,
Software defined network, networks and telecommunications.

Najib Elkamoun is a professor at Faculty of sciences, Chouaib Doukkali University, El Jadida,

Morocco. He is a researcher member of STIC laboratory and header of Network and
Telecommunications team. His research interest includes: NGN, MPLS, Networks, QoS in
mobile networks, wireless networks.

Fatima Lakrami is a researcher and a professor at Science Faculty, Chouaîb Doukkali

University, El Jadida, Morocco. She got her Doctorate in Telecommunication and Networking in
2014 from Chouaîb Doukkali University, El Jadida, Morocco. Her research interests cover
wireless networks performance evaluation, VANETs, Security.

A new SDN-based load balancing algorithm for IoT devices (Hind Sounni)

View publication stats