An efficient reconfigurable workload balancing scheme for fog computing network using internet of things devices

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 13, No. 6, December 2023, pp. 6525~6537
ISSN: 2088-8708, DOI: 10.11591/ijece.v13i6.pp6525-6537  6525
Journal homepage: http://ijece.iaescore.com
An efficient reconfigurable workload balancing scheme for fog
computing network using internet of things devices
Nandini Gowda Puttaswamy1
, Anitha Narasimha Murthy2
1
Computer Science and Engineering, East Point College of Engineering and Technology, Visvesvaraya Technological University,
Belagavi, India
2
Computer Science and Engineering, BNM Institute of Technology, Visvesvaraya Technological University, Belagavi, India
Article Info ABSTRACT
Article history:
Received Jun 5, 2023
Revised Jul 11, 2023
Accepted Jul 18, 2023
Nowadays a huge amount of data has been communicated using fog nodes
spread throughout smarty cities. the communication process is performed
using fog nodes which are co-located with cellular base stations (BSs) that
can move the computing resources close to internet of things (IoT) devices.
In smart cities, a different type of data flow has been communicated through
IoT devices. The communication process performs efficiently using the
remote cloud. The IoT devices very close to the BS can communicate data
without using fog nodes. Due to these phenomena, workload unbalancing
occurs in IoT devices communicating in fog computing networks. Hence, it
generates communication and computing latency. The task distribution
process between the IoT devices is unbalanced. Hence, congestion and loss
of information occur in fog computing network. A proposed reconfigurable
load balancing algorithm (RLBA) is efficiently balancing the workload by
reconfigurable communication channels and deviates the task with respect to
the BS locations, IoT devices density and load IoT devices in each fog nodes
in a network to minimize the communication and computing latency. As per
the performance analysis, the proposed algorithm shows better performance
as compared to conventional methods’ average latency ratio, communication
latency ratio, computing load and traffic load.
Keywords:
Base station
Fog node computing network
Fuzzy golden eagle load
balancing
Internet of things
Reconfigurable workload
balancing algorithm
Work load allocation
This is an open access article under the CC BY-SA license.
Corresponding Author:
Nandini Gowda Puttaswamy
Computer Science and Engineering, East Point College of Engineering and Technology, Visvesvaraya
Technological University
Belagavi-590018, India
Email: nandini.educator@gmail.com
1. INTRODUCTION
The development of smart gadgets that can perceive the environment physical information has
received a lot of research attention in recent years [1]. The term “internet of things (IoT)” refers to a concept
in which linked the smart devices of one another over the internet and equipped with data analytics. IoT is
currently employed in many different applications, including smart cities, smart homes, smart transportation,
and smart health [2]. As the number and types of devices connected to the IoT increases, managing these
devices efficiently becomes a vital consideration. As the number and variety of devices increase, so does the
need for cloud-based thinking [3], [4]. But since the data streams produced by IoT devices are sent to a
distant cloud over the internet, it uses an amount of energy and core network bandwidth [5].
Many delay-sensitive IoT applications typically have a significant data streams processing with a
delay in cloud hosted architectures where the main network is frequently located long distance from the IoT
devices [6]. Fog nodes, which bring computer resources closer to devices of IoT and IoT users, are one

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 13, No. 6, December 2023: 6525-6537
6526
option for easing the traffic burden on the core network and lowering IoT devices latency. Sometimes IoT
application require a huge number of linked devices that produce enormous volumes of data. A paradigm
called fog computing seeks to deal with the difficulties of processing this data fast and effectively. Fog
nodes, often referred to as fog servers or fog devices, are essential to fog computing because they bring
storage and processing resources closer to the network’s edge [7], [8].
IoT devices sense data flows, which are transmitted to the appropriate base stations (BSs) in a fog
network processed by fog nodes that are physically close to the BSs. Therefore, each data flow’s latency is
the sum of the communications times between an IoT device and a BS and a fog node. In the IoT, there are
two different types of latency in communications: device latency and channel latency. The delay brought on
by the wireless channel is referred to as channel latency. In a network, for example, when a wireless
transmission involves multiple BSs, the wireless signal has to be transmitted from one BS to another
sequentially. As a result, there will be significant communication delay due to the time wasted by round-trip
communications. This kind of delays is known as channel latency. On the other hand, an IoT device will
experience a delay if it sends its data through fog nodes, which are dispersed in areas with less traffic. This is
because data transit through wired or wireless lines takes time. Device latency is the term for this delay.
Therefore, it is required for the data flows total latency in a network that computational tasks are dynamically
distributed among fog nodes. The fog node workload is closely correlated with the number of IoT devices
connected to its corresponding BS because each fog node in this article is coupled to a single BS. That is,
data flows from any IoT device that joins an IoT network are offloaded to a single co-located fog node [9].
Research has shown that nearby wireless service blocks BSs overlap and give coverage is as shown
in Figure 1, which depicts the basic architecture computing of fog network using IoT devices. To balance
loads among these BSs, fog nodes placed within these overlapping BSs might be connected to appropriate
BSs. This connection has a significant influence on both the computation demands of traffic and fog nodes of
BSs. Load balancing should take into account both the traffic loads of BSs and the computation loads of fog
nodes in order to lower the latency of all data flows among devices and servers in the network. This is
because latency is made up of two components: communication delay and calculation latency. In essence,
when overloaded some BSs, they turn into bottlenecks and slow down communications for connected
devices. To lessen their traffic demands, some IoT devices connected to these overcrowded BSs should be
offloaded to nearby BSs.
Our aim is to link devices of IoT to various fog and BS fog network nodes to lower the latency of all
network data flows in the IoT devices. The average-latency ratios at BSs/fog nodes are used to represent,
respectively, the processing latency in fog network nodes and the communications delay in BSs. We
introduce the models for the compute demands at fog network nodes and mean traffic loads at BSs/fog nodes.
Additionally, we have examined the effect load balancing has on the usual latency of IoT device traffic [10],
[11]. As a solution to the problem of load-balancing in a fog IoT device network, we create a distributed IoT
device association algorithm to map IoT based network devices to pertinent BSs/fog nodes, thereby reducing
the mean latency of all network IoT data flows. The LAB accomplishes this by employing a greedy heuristic
to repeatedly estimate traffic and resource burdens per BS before broadcasting this information back to
devices of IoT so they may select practical and appropriate destinations. Additionally, a simulation is offered
to demonstrate the convergence.
Figure 1. The fundamental structure of fog computing network using IoT device
2. RELATED WORKS
Since fog computing is close to users and devices off IoT, in some studies have concentrated on
merging IoT with fog network computing. Fog computing has applications that were developed by
Yang et al. [12]. In the environment of networking, Marbukh [13] outlined the advantages and disadvantages
of fog computing and stated that it can cover the gaps left by IoT. An architectural strategy (EdgeIoT) was

Int J Elec & Comp Eng ISSN: 2088-8708 
An efficient reconfigurable workload balancing scheme for fog computing … (Nandini Gowda Puttaswamy)
6527
suggested by Donno and Dragoni [14] to control the IoT network device data streams at the fog network
node. To further optimize traffic data flows and IoT network resources, Ju et al. [15] presented adaptive fog
computing systems for IoT networking. It has been suggested that fog computing will enhance the
functionality of mobile and IoT apps. In this work, we investigate the association strategy, the FogNode
allocation technique, and the placement of virtual machines based on traffic load balancing and quality of
service (QoS) in the medical cyber-physical system. Task scheduling and picture placement in a software-
defined embedded system have been presented by Tadesse et al. [16] in fog computing as a way to reduce
task request response times. A workload placement approach was developed by Kuzman et al. [17] in order
to optimize the response times of hierarchical edge cloud network all tasks. The algorithm distributes tasks
among various tiers of fog network nodes and allots computational resources from each fog network node for
the tasks that each fog network node is given. In order to improve user task response times and lower cloudlet
brown energy consumption, Elavarasan and Vincent [18] introduced a workload allocation technique, known
as workload allocation (WALL), in a hierarchical cloudlet network. Tomar et al. [19] suggested transferring
mobile users’ virtual-machines (VM) among scattered cloudlets while also taking into account the energy
requirements for VM migrations in order to decrease the overall consumption of energy dispersed cloudlets.
The workload allocation method described in this study allocates cloudlet users to edge computing resources
at various sites and then schedules tasks for the edge computing resources in accordance with the real-time
needs of the associated users. Some studies also look at how wireline delay affects user task latency when
edge computing resources are set up at various locations. Keep in mind that all of the aforementioned works
only take wired communications latency into account, leaving out wireless delay. To reduce user response
times for mobile edge computing networks, Hu et al. [20] present a model for load balancing cloudlets. In the
present research, it is assumed that each user’s wireless delay is constant. Some studies suggested adjusting
BS transmission power to modify data speeds and shorten user wireless delays. Additionally, a number of
research on cellular mobile networks’ energy efficiency have been done.
3. SYSTEM MODEL
A fog IoT based network topology with fog nodes connected to BSs and overlapped coverage zones
between close-by BSs is shown in Figure 2. It should be mentioned that all BSs use the next generation IoT
(NBIoT) interface to deliver to IoT network devices communications services [21]. Because of the
distribution of a workload across fog network nodes, the data flows require to pass over the mobile cellular,
the IoT information flows are often to be processed at the local BS fog network node is preferred, which adds
additional latency. On the other hand, to collect data a central controller is required on both the workload
devices of IoT in the network topology and fog nodes in order to implement a centralized algorithm for the
real-time allocation of workload among fog network nodes. For big size networks, such metropolitan area
networks, this complexity makes it unusable. Therefore, we consider that the fog network node connected to
the IoT network device’s BS processes network data flows from IoT devices rather than other fog nodes.
Other recent studies, like [22], similarly make the same assumption based on similar concerns. Because from
network no need to collect data from every device, it should be noted that in this situation the compute loads
can be distributed easily and with flexibility over several fog nodes.
Figure 2. Fog network architecture

 ISSN: 2088-8708
6528
4. TRAFFIC LOAD MODEL
In this study, we considered that traffic loads arrive at a location x averaging a unit area rate, 𝛾(𝑥)
as by using a poisson-point process. Additionally, we suppose that each BS has a unique fog node, F.
Indicate x as a place inside C and A as the total area covered by all BSs. We assume that IoT network data
flows arrive at location x at a poisson-point process average rate of 𝛾(𝑥) per unit area. Traffic loads vary
spatially as a result of inhomogeneity. The main notations used in this work is shown in Table 1. The uplink
channel gain from an IoT device’s position to BS j, the noise power 𝜃2
, and the IoT device transmission
power at that location are all specified as 𝑢𝑗(𝑥). The IoT device-based network signal-to-noise ratio (SNR)
may be computed using (1) and (2) illustrates in the theorem of Shannon Hartley, if we denote 𝑏𝑗(𝑥) as the
IoT transmission power at location x and it is logarithmic function is denoted by 𝜆𝑗(𝑥).
ℶ𝑗(𝑥) =
𝑇(𝑥)𝑢𝑗(𝑥)
𝜃2 (1)
𝑏𝑗(𝑥) = 𝜑𝑗 log(1 + 𝜆𝑗(𝑥)) (2)
where 𝑏𝑗(𝑥) is represented IoT devices load traffic of average and 𝜑𝑗 is described the jth
BS total bandwidth
[23]. In various network environments, fog network nodes can be deployed by grouping smart devices
together and creating a group of networks known as fog networks. The fog topology greatly improves the
coverage of wireless network access points that are fixed to particular locations. In this arrangement, data
flows received up by an IoT device are sent to its BS, where the fog network nodes is processed residing next
to the BS. We will therefore concentrate on IoT device uplink connections and data processing of fog node in
view to determine the flow of data latency. Equation (3) has been used to calculate the network load density
average traffic for IoT devices.
∝𝑗 (𝑥) =
𝛾(𝑥)𝑚(𝑥)𝜌𝑗(𝑥)
𝑏𝑗(𝑥)
(3)
where 𝜌𝑗(𝑥) is represented as binary variable with the location of ‘x’ is associated fog network nodes with
the jth
BS and ℵ𝑗 is described as avaearge traffic at BS, 𝑚(𝑥) is identified as exponential data flow
distribution average value and the IoT device’s typical traffic-load density at location x in BS j. The fog node
average traffic has been calculated using (4).
ℵ𝑗 = ∑ ∝𝐽 (𝑥)
𝑥𝜖𝐶 (4)
A several factors, including user fairness and network capacity are considered to manage the IoT
network devices access radio resources in mobile communications. In this study, IoT network devices are
arranged in round-robin fashion for uplink transmissions such that uplink channel access by several devices
concurrently for the sake of analytical tractability. While the network data rates at each place are provided,
the IoT based network traffic arrival information rate at location ‘x’ also complies with the poisson-process,
the information traffic volumes of data flows follow the exponential distribution, and the service duration of
IoT network data flows at location ‘x’ also exponential distribution satisfied [24] and the mean service time
(MST) of network data flows at point x is determined by (5).
∅𝑗(𝑥) =
𝑚(𝑥)
𝑏𝑗(𝑥)
(5)
As a simulation result, an M/M/1-processor serves sharing queue as the foundation for a BSs uplink
communications (BS) [25]–[29]. As per the suggested approach, different IoT network devices will equitably
information share the average ratio energy and IoT network devices information resources of a BS even when
they have different IoT data rates depending on their channel conditions. Furthermore, we must constantly
ensure that ℵ𝑗 is less than one in order to maintain the queue’s stability. Processor sharing given the M/M/1-
network data queue of a BS, mean value of delivery time at location x is given by (6).
𝛽𝑗(𝑥) =
𝑚(𝑥)
𝑏𝑗(𝑥)(1−𝜂𝑗)
(6)
The mean waiting time for each data flow at point x is given by (7).

6529
𝜑𝑗 = 𝛽𝑗(𝑥) − ∅𝑗(𝑥) =
𝜂𝑗𝑚(𝑥)
𝑏𝑗(𝑥)(1−𝜂𝑗)
(7)
The symbol 𝜗𝑗(𝑥) represents the ratio of latency time of waiting to service in BS j for IoT network
data flows at the location x. Then, since all of the devices in IoT connected to BS j have the same delay ratio
and 𝜗𝑗(𝑥) only depends on the network traffic load of BS j, we define BS j’s latency ratio communications as
we can deduct from (8) that raising the network traffic load ρj of BS j will increase j. IoT devices connected
to BS µj must wait longer to reach the channel of transmission when µj>0, which is consistent with their
delay of average delivery as indicated by (8).
𝜗𝑗 =
𝜂𝑗
1−𝜂𝑗
(8)
5. PROBLEM FORMULATION
In this study, we balance workloads between fog and BSs nodes in an effort to reduce the data flows
latency. We write 𝜏(𝛼) = ∑ 𝜗𝑗 + 𝜗𝑗
̂
𝑗∈𝐹 as the fog network latency ratio. The goal of the suggested technique
is to determine the ideal load distribution among BSs in view to reduce the fog network’s latency ratio.
Consequently, the issue might be stated as (9) to (13).
𝑇1: min
𝛼
∁(𝛼) (9)
∑ 𝜗𝑗(𝑥) = 1,
𝛼
𝑗∈𝐹 ∀𝑥∈ 𝜏 (10)
0 ≤ 𝜂𝑗 ≤ 𝜂𝑚𝑎𝑥, ∀𝑗∈ 𝐹 (11)
0 ≤ 𝜂𝑗
̂ ≤ 𝜂̂𝑚𝑎𝑥, ∀𝑗∈ 𝐹 (12)
𝜂𝑗(𝑥)𝜖{0,1}, ∀𝑥∈ 𝜏 (13)
There can only be one BS associated with each location, according to constraints (10) and (13). The
computation amount at fog node i is constrained by constraint (12). In load balancing, two factors influence
the computation latency of data flows: traffic loads and computing loads. In an IoT network, when
bandwidth-constrained devices are deployed in some areas, data congestion on overloaded BSs may result in
a high latency for data flows [30]–[33]. On the other side, in a fog network, some fog nodes may become
overloaded due to the heavy computational loads it has to host.
6. LAB: A DISTRIBUTED IoT DEVICE ASSOCIATION SCHEME
In this work, we suggest a load-shaping mechanism for wireless IoT networks that are delay-
tolerant. The controller in our suggested system continuously calculates the compute and fog nodes (BSs)
traffic loads, then broadcasts the results to devices of IoT [34]. The conclusion is that IoT devices can choose
BSs iteratively based on information of real load and their data rates of uplink to various BSs. We
demonstrate that this distributed system can manage problems like service disruptions while achieving the
best possible trade-off between delays and power consumption.
6.1. Algorithm for IoT device side
Initially the kth iteration has been considered for all traffic loads 𝜂𝑗 and manuplated loads 𝜂̂𝑗 to
devices of IoT. Based on fog node network 𝑀(𝛼) as shown in (14).
𝜕𝑀(𝛼)
𝜕𝛼𝑗(𝑥)
= 𝛾(𝑥)
𝑒𝑗(𝑥)𝑚(𝑥)(1−𝜂
̂𝑗(𝑘))2+𝑏𝑗(𝑥)𝑣(𝑥)(1−𝜂𝑗(𝑘))2
𝑒𝑗𝑏𝑗(𝑥)(1−𝜂
̂𝑗(𝑘))2(1−𝜂𝑗(𝑘))2 (14)
IoT devices have been selected based on broadcast messages. Determine the BS is by using (15),
𝑡𝑘(𝑥) = 𝑎𝑟𝑔 max
𝑗∈𝐹
𝑒𝑗(𝑥)𝑏𝑗(𝑥)𝜓𝑗(𝑘) (15)

 ISSN: 2088-8708
6530
where 𝑒𝑗(𝑥) is represented capacity computing (in CPU cycle/sec) for fog node ‘J’ and 𝜓𝑗(𝑘) is described
the in 16 and 𝑡𝑘(𝑥) is identified as BS index and it controlled by condition is given in (17).
𝜓𝑗(𝑘) =
(1−𝜂
̂𝑗(𝑘))2(1−𝜂𝑗(𝑘))2
𝑒𝑗(𝑥)𝑚(𝑥)(1−𝜂
̂𝑗(𝑘))2+𝑏𝑗(𝑥)𝑣(𝑥)(1−𝜂𝑗(𝑘))2 (16)
𝛼𝑗
𝑘
(𝑥) = {
1, 𝑖𝑓 𝑗 = 𝑡𝑘(𝑥) , ∀𝑥∈ 𝜏
0, 𝑖𝑓 𝑗 ≠ 𝑡𝑘(𝑥) , ∀𝑥∈ 𝜏
} (17)
6.2. Algorithm for BS side
A BS must be aware of both its own traffic load and the compute load of the relevant fog node in
order to estimate the intermediary IoT association. For each IoT device in the iteration, it must therefore
estimate an intermediate IoT association. Then, using their side algorithm, IoT devices choose their BSs or
fog nodes depending on the predicted load information among BSs, and the current IoT device association in
the kth
iteration changes to 𝜂𝑗(𝑘) as a result. As a result, BS ‘j’ can assessment IoT association an
intermediate for location ‘x’ in the following iteration using the intermediate 𝛼𝑗
~𝑘
(𝑥) (estimated by a BS) and
the current IoT device association 𝛼𝑗
𝑘
(𝑥) (determined by IoT devices) equations:
𝛼𝐽
~𝐾+1(𝑥) = (1 − £)𝛼𝑗
𝑘
(𝑥) + £𝛼𝑗
~𝑘
(𝑥) (18)
where £ is represented system parameter with a range 0 ≤ £ ≤ 1, if devices of IoT are associate with (k+1),
then the traffic load has been calculated by the (19).
𝜂𝑗(𝑘 + 1) = ∫
𝛾(𝑥)𝑚(𝑥)𝛼𝑗
~𝑘+1
(𝑥)
𝑏𝑗(𝑥)
𝑥∈𝜏
𝑑𝑥 (19)
Similarly, the next computing traffic load at fog node ‘j’ is obtained from (20).
𝜂𝑗
̂ (𝑘 + 1) = ∫
𝛾(𝑥)𝑣(𝑥)𝛼𝑗
~𝑘+1
(𝑥)
𝑒𝑗(𝑥)
𝑥∈𝜏
𝑑𝑥 (20)
The BS selected algorithm has been done by using algorithm 1.
Algorithm 1. The BS side algorithm
Input: BS choice for IoT devices:𝑡𝑘(𝑥), ∀𝑥∈ 𝜏. The kth iteration’s intermediate IoT device association vector, 𝛼𝑗
𝑘
.
Output: The estimated traffic loads of BSs 𝜂 (𝑘 + 1) and the computing loads is estimated of fog nodes 𝜂̂(𝑘 + 1) in the (k+1) th.
The estimated loads computing of fog nodes 𝜂̂(𝑘 + 1) in the (𝑘 + 1) th and the estimated traffic loads of BSs in the 𝜂 (𝑘 + 1) th.
a. Update the IoT device association intermediate for different locations based on: 𝛼𝐽
~𝐾+1(𝑥) = (1 − £)𝛼𝑗
𝑘(𝑥) + £𝛼𝑗
~𝑘(𝑥), 𝑥 ∈ 𝜏, 𝑗 ∈ 𝐹.
b. Calculate 𝜂𝑗(𝑘 + 1) and 𝜂̂𝑗(𝑘 + 1) based on (19) and (20);
c. Return 𝜂𝑗(𝑘) and 𝜂̂𝑗(𝑘 + 1).
Treated in algorithm 1. Equation (21) contains the feasible set for problem P1, as is well known.
𝑀
̂ = {𝛼|𝜂𝑗 = ∫
𝛾(𝑥)𝑚(𝑥)𝛼𝑗(𝑥)
𝑏𝑗(𝑥)
𝑥∈𝜏
𝑑𝑥, } (21)
where 𝛼𝑗(𝑥) ∈ {0,1}, 0 ≤ 𝜂𝑗 ≤ 𝜂𝑚𝑎𝑥 and ∑ 𝛼𝑗(𝑥) = 1, ∀𝑗 ∈ 𝐹,
𝑗∈𝐹 ∀𝑥∈ 𝜏
Treated in algorithm 1. As we are aware, the equation (22) can be used to define the feasible set of
problem P1.
𝑀 = {𝛼|𝜂𝑗 = ∫
𝛾(𝑥)𝑚(𝑥)𝛼𝑗(𝑥)
𝑏𝑗
𝑥∈𝜏
𝑑𝑥}
𝛼𝑗(𝑥) ∈ {0,1}, 0 ≤ 𝜂𝑗 ≤ 𝜂𝑚𝑎𝑥
∑ 𝛼𝑗(𝑥) = 1, ∀𝑗 ∈ 𝐹,
𝑗∈𝐹 ∀𝑥∈ 𝜏 (22)
We must employ IoT association an intermediate value to lower the average latency ratio 𝐿() in
each iteration because the traffic load (or computing load) vector for problem P1 is not convex, as indicated
by the 𝛼𝑗(𝑥) ∈ {0,1}. We demonstrate that the load vectors computing and traffic load will converge to be in

6531
the feasible set by first relaxing to make 0≤𝛼𝑘≤1. Given in (23), the feasible set relaxed for problem P1 can
be expressed.
𝑀
̂ = {α|ηj = ∫
γ(x)m(x)αj(x)
bj(x)
x∈τ
dx, }
𝛼𝑗(𝑥) ∈ {0,1}, 0 ≤ 𝜂𝑗 ≤ 𝜂𝑚𝑎𝑥
∑ 𝛼𝑗(𝑥) = 1, ∀𝑗 ∈ 𝐹,
𝑗∈𝐹 ∀𝑥∈ 𝜏 (23)
Theorem 1: The feasible relaxed set 𝑀
̂ is a convex set.
Proof: As the set 𝑀
̂ includes any convex combination of α, it is defined in 𝑀
̂.
Theorem 2: The convex function of α, when is defined by objective function (𝛼), when ‘α’ is defined in 𝑀
̂.
Proof: ∇2
𝜏(𝛼) > 0 during (𝛼) is defined in 𝑀
̂ proved by showing this equation.
6.3. Analysis of the algorithm
In this part, we will examine the LAB scheme’s convergence and optimality with regard to problem P1.
Theorem 3: Gives a direction for 𝜏(𝛼
̂) at 𝛼
̂k
, when 𝛼
̂k+1
≠ 𝛼
̂k
, 𝛼
̂k+1
.
Proof: As 0 ≤ 𝛼
̂𝑗
𝑘
(𝑥) ≤ 1, 𝜏(𝛼
̂) is defined in 𝑀
̂, As shown in Theorem 2 𝜏(𝛼
̂) is a convex function of 𝛼
̂,
and thus, we need to prove 〈∇∁(𝛼
̃𝑘), 𝛼 ̃𝑘+1
− 𝛼
̃𝑘〉 < 0 thus, we have (24).
〈∇∁(𝛼
̃𝑘), 𝛼 ̃𝑘+1
− 𝛼
̃𝑘〉
= ∫ ∑ 𝛾(𝑥)𝑣(𝑥)
𝛼𝐽
~𝐾+1(𝑥)−𝛼𝐽
~𝐾(𝑥)
𝑒𝑗𝑏𝑗(𝑥)𝜓𝑗(𝑘)
𝑗 ∈𝐹
𝑥∈𝜏
= ∫ 𝛾(𝑥)𝑣(𝑥)
𝛼𝐽
~𝐾+1(𝑥)−𝛼𝐽
~𝐾(𝑥)
𝑥∈𝜏
(24)
Based on (18), we can compute using (25).
𝛼𝐽
~𝐾+1(𝑥) − 𝛼𝐽
~𝐾(𝑥) = (1 − £)(𝛼𝑗
𝑘
(𝑥) + £𝛼𝑗
~𝑘
(𝑥)) (25)
As we know that, in (26).
𝛼𝐽
~𝐾(𝑥) = {
1, 𝑖𝑓 𝑗 = 𝑡𝑘(𝑥)
0, 𝑖𝑓 𝑗 = 𝑡𝑘(𝑥)
} (26)
The kth
iteration is adopted for BS selection rule at the destine side, then it can be achieved by the equations
𝑡𝑘(𝑥) = 𝑎𝑟𝑔 max
𝑗∈𝐹
𝑒𝑗(𝑥)𝑏𝑗(𝑥)𝜓𝑗(𝑘), from we can derive the (27) and (28):
∑ (1 − £)
𝛼𝑗
𝑘(𝑥)−𝛼𝐽
~𝐾(𝑥)
𝑗∈𝐹 ≤ 0 (27)
𝑠𝑖𝑛𝑐𝑒 𝛼𝐽
~𝐾+1(𝑥) ≠ 𝛼𝐽
~𝐾(𝑥)
∑ (1 − £)
𝛼𝑗
𝑘(𝑥)−𝛼𝐽
~𝐾(𝑥)
𝑗∈𝐹 < 0 (28)
Thus, we have proved 〈∇∁(𝛼
̃𝑘), 𝛼 ̃𝑘+1
− 𝛼
̃𝑘〉 < 0.
By proving the subsequent theorem as part of the LAB scheme, we will also examine if the optimal
selection rule for BS each iteration at the IoT device side is the best choice.
Theorem 1: The ideal IoT device association rule at the fog node side is the following given the advertised
traffic loads of BSs and computing loads of fog nodes 𝑡𝑘(𝑥) = 𝑎𝑟𝑔 max
𝑗∈𝐹
𝑒𝑗(𝑥)𝑏𝑗(𝑥)𝜓𝑗(𝑘), where x is a
feature vector associated with an IoT device.
Proof: The proposed kth iteration side algorithm for IoT side algorithm achieves association IoT device given
by 𝑡𝑘(𝑥) = 𝑎𝑟𝑔 max
𝑗∈𝐹
𝑒𝑗(𝑥)𝑏𝑗(𝑥)𝜓𝑗(𝑘) Meanwhile, let ‘𝛼’ denote any other possible IoT devices, with a
condition and (29).

 ISSN: 2088-8708
6532
< ∇∁(𝛼𝑘), 𝛼′
− 𝛼𝑘
> 0
< ∇∁(𝛼𝑘), 𝛼′
− 𝛼𝑘
>
= ∫ ∑ 𝛾(𝑥)𝑣(𝑥) (𝛼′
− 𝛼𝑗
𝑘
(𝑥))
1
𝑑𝑥
𝑗 ∈𝐹
𝑥∈𝜏
= ∫ 𝛾(𝑥)𝑣(𝑥) ∑ (𝛼′
− 𝛼𝑗
𝑘
(𝑥))
𝑗∈𝐹
1
𝑥∈𝜏
𝑑𝑥 (29)
since (30)
𝑡𝑘(𝑥) = 𝑎𝑟𝑔 min
𝑗∈𝐹
𝑒𝑗(𝑥)𝑏𝑗(𝑥)𝜓𝑗(𝑘)
𝛼𝑗
𝑘
(𝑥) = {
1, 𝑖𝑓 𝑗 = 𝑡𝑘(𝑥) , ∀𝑥∈ 𝜏
0, 𝑖𝑓 𝑗 ≠ 𝑡𝑘(𝑥) , ∀𝑥∈ 𝜏
} (30)
then, we have
∑ 𝛼𝑗
′
(𝑥)
1
≥ ∑ 𝛼𝑗
𝑘
(𝑥)
1
𝑗∈𝐹
𝑗 ∈𝐹
hence, 〈∇∁(𝛼𝑘), 𝛼′
− 𝛼𝑘 〉 > 0. Therefore, 𝛼𝑘
is an optimal IoT device association in the kth
iteration.
7. NUMERICAL RESULTS
In this section, we present simulations to gauge how well our proposed approach works. In this part,
we compare it to the Best SNR algorithm and distributed algorithm. The fundamental goal of the distributed
algorithm is to divide workloads of traffic among BSs as efficiently as possible in view to reduce
communications delay ratio (i.e., ∑ 𝜗𝑗
𝑗∈𝐹 ) without taking load distribution at fog nodes into account. The
Best SNR algorithm, on the other hand, aims to link IoT devices with BSs that offer the conditions best
channel. The Table 1 displays the simulation parameters for a comparison of proposed and standard methods’
performance.
Table 1. Simulation parameter
SI.NO Parameter Range
1. Number of BSs 6
2. Total network area 3,000×2,000 m2
3. Number of areas divided 15,000 Location
4. Location area 20×20 m
5. Mean arrival rate per unit area is set 0.50 flows/second
6. Mean traffic area 0.05 Mbits
Figure 3 shows the network topology BSs distribution with respect to the parameter consider in
Table 1. The Figure 4 shown the performance analysis between proposed method and conventional methods
average latency ratio 𝐿(𝛼) with respect to the 𝛾 = 0.6, 𝑒𝑗 = 7.6 ∗ 106
. The average delay ratio 𝐿() for the
various algorithms is shown in Figure 5. Figure 6 displays the communication delay (average) ratio in
relation to the number of iterations, 𝛾 = 0.6, 𝑒𝑗 = 7.6 ∗ 106
, whereas Figures 7 and 8 display the compute
loads placed on various fog nodes within a network structure.
Figure 9 shows the average delay ratio in terms of each fog’s capacity (𝛾 = 0.5), and Figure 10
compares the performance of the suggested method to that of the more traditional approach in terms of the
flow arrival rate (𝑥), where 𝛾 (𝑥) , 𝑒𝑗 = 7.6 ∗ 106
. As a result, as compared to fuzzy golden eagle load
balancing (FGELB), the compute load balancing of reconfigurable load balancing algorithm (RLBA) is
unable to reduce average latency. However, as the average traffic arrival rate rises, both the traffic loads and
compute load in a network become heavy, which causes the average latency ratio of RLBA to develop slowly
while the performance of the FGELB declines quickly. In the present scenario, average latency is impacted
by both traffic loads among BSs and computing loads among fog nodes. However, this method does not
concentrate on balancing traffic loads between BSs but simply balances compute loads across fog nodes. As
RLB takes into account both load balancing algorithms, it can still maintain low average latency in spite of
large computing loads. This indicates that some fog nodes are crowded, particularly when the network is
experiencing heavy compute load.

6533
Figure 3. Fundamental base station distribution in network topology
Figure 4. The average latency ratio 𝐿(𝛼) performance analysis between proposed and conventional methods
Figure 5. The average latency ratio 𝐿(𝛼) for the different algorithms with respect to the 𝛾 = 0.6, 𝑒𝑗 = 7.6 ∗ 106

 ISSN: 2088-8708
6534
Figure 6. The average latency ratio 𝐿(𝛼) performance analysis between proposed and conventional methods
Figure 7. Computing loads of different fog nodes
Figure 8. Various BSs traffic loads
2 4 6
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
Computing
Load
Variours BSs
RLBA
FGELB
Best SNR
2 4 6
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Traffic
Load
Variours BSs
RLBA
FGELB
Best SNR

6535
Figure 9. The performance analysis between proposed and conventional methods with respect to
𝛾 = 0.6 , 𝑒𝑗 = 7.6 ∗ 106
Figure 10. The performance analysis between proposed and conventional methods with respect to
𝛾 = 0.6 , 𝑒𝑗 = 7.6 ∗ 106
8. CONCLUSION
In this paper, the main concentration is on the proper workload distribution in smart cities using a
RLBA. When a huge number of data flows across the IoT device in a fog computing network, there may be a
loss of information, more attenuation, and unbalancing task in fog nodes. to overcome the above-mentioned
drawback, a proposed RLBA has been used to optimize the workload by considering a few standard
parameters in the fog computing network. With respect to the average latency ratio, 𝐿(𝛼) performance
analysis has been done by considering (𝛾 = 0.6, 𝑒𝑗 = 7.6 ∗ 106
.) the proposed RLBA. The result shows
better performance as compared to the conventional method of about 1.6% with respect to the
communication latency ratio. As per the analysis of performance that has been done by considering
(𝛾 = 0.6, 𝑒𝑗 = 7.6 ∗ 106
), the proposed RLBA shows better performance as compared to the conventional
method of about 0.96%. With respect to the computing load, the proposed method shows better performance
of about 0.26% and 1.21% as compared to FGELB and best SNR respectively. With respect to the traffic
load, the proposed method RLBA optimized to 0.98% and 1.25% as compared to FGELB and best SNR
respectively.

 ISSN: 2088-8708
6536
FUTURE SCOPE
The emerging technology is six generation (6G), our proposed method is capable to operate the 6G
communication in a fog computing network using IoT devices without lose of information and attention.
ACKNOWLEDGEMENTS
The author would like to thank East Point College of Engineering and Technology, Sapthagiri
College of Engineering, BNM Institute of Technology, Visvesvaraya Technological University (VTU),
Belagavi for all the support and encouragement provided by them to take up this research work and publish
this paper.
REFERENCES
[1] Z. Liu, Y. Yang, Y. Chen, K. Li, Z. Li, and X. Luo, “A multi-tier cost model for effective user scheduling in fog computing
networks,” in IEEE INFOCOM 2019-EEE Conference on Computer Communications Workshops (INFOCOM WKSHPS), Apr.
2019, pp. 1–6, doi: 10.1109/INFCOMW.2019.8845252.
[2] S. Thazeen, S. Mallikarjunaswamy, and M. N. Saqhib, “Septennial adaptive beamforming algorithm,” in 2022
International Conference on Smart Information Systems and Technologies (SIST), Apr. 2022, pp. 1–4, doi:
10.1109/SIST54437.2022.9945753.
[3] B. Jia, H. Hu, Y. Zeng, T. Xu, and Y. Yang, “Double-matching resource allocation strategy in fog computing networks based on
cost efficiency,” Journal of Communications and Networks, vol. 20, no. 3, pp. 237–246, Jun. 2018, doi:
10.1109/JCN.2018.000036.
[4] S. Mallikarjunaswamy, N. M. Basavaraju, N. Sharmila, M. H. N, S. Pooja, and B. L. Deepak, “An efficient big data gathering in
wireless sensor network using reconfigurable node distribution algorithm,” in 2022 Fourth International Conference on Cognitive
Computing and Information Processing (CCIP), Dec. 2022, pp. 1–6, doi: 10.1109/CCIP57447.2022.10058620.
[5] C. Lin, G. Han, X. Qi, M. Guizani, and L. Shu, “A distributed mobile fog computing scheme for mobile delay-sensitive
applications in SDN-enabled vehicular networks,” IEEE Transactions on Vehicular Technology, vol. 69, no. 5, pp. 5481–5493,
May 2020, doi: 10.1109/TVT.2020.2980934.
[6] Q. Jin, R. Lin, H. Zou, and F. Yang, “A distributed fog computing architecture supporting multiple migrating mode,” in 2018 5th
IEEE International Conference on Cyber Security and Cloud Computing (CSCloud)/2018 4th IEEE International Conference on
Edge Computing and Scalable Cloud (EdgeCom), Jun. 2018, pp. 218–223, doi: 10.1109/CSCloud/EdgeCom.2018.00046.
[7] F. M. Siddiqui, “Fog computing application for effective fronthaul management in fifth generation networks,” in 2018
International Conference on Computing, Electronics and Communications Engineering (iCCECE), Aug. 2018, pp. 107–112, doi:
10.1109/iCCECOME.2018.8658911.
[8] S. Jin, Z. Zhu, Y. Yang, M.-T. Zhou, and X. Luo, “Alternate distributed allocation of time reuse patterns in Fog-enabled
cooperative D2D networks,” in 2017 IEEE Fog World Congress (FWC), Oct. 2017, pp. 1–6, doi: 10.1109/FWC.2017.8368532.
[9] S. Pooja, S. Mallikarjunaswamy, and N. Sharmila, “Image region driven prior selection for image deblurring,” Multimedia Tools
and Applications, vol. 82, no. 16, pp. 24181–24202, Jul. 2023, doi: 10.1007/s11042-023-14335-y.
[10] S. Kitanov and T. Janevski, “Energy efficiency of fog computing and networking services in 5G networks,” in IEEE EUROCON
2017-17th International Conference on Smart Technologies, Jul. 2017, pp. 491–494, doi: 10.1109/EUROCON.2017.8011159.
[11] L. Velasco and M. Ruiz, “Flexible fog computing and telecom architecture for 5G networks,” in 2018 20th International
Conference on Transparent Optical Networks (ICTON), Jul. 2018, pp. 1–4, doi: 10.1109/ICTON.2018.8473934.
[12] L. Yang, H. Zhu, H. Wang, H. Qian, and Y. Yang, “Incentive propagation mechanism of computation offloading in fog-enabled
D2D networks,” in 2018 IEEE 23rd International Conference on Digital Signal Processing (DSP), Nov. 2018, pp. 1–4, doi:
10.1109/ICDSP.2018.8631842.
[13] V. Marbukh, “Towards fog network utility maximization (FoNUM) for managing fog computing resources,” in 2019 IEEE
International Conference on Fog Computing (ICFC), Jun. 2019, pp. 195–200, doi: 10.1109/ICFC.2019.00032.
[14] M. De Donno and N. Dragoni, “Combining AntibIoTic with fog computing: AntibIoTic 2.0,” in 2019 IEEE 3rd International
Conference on Fog and Edge Computing (ICFEC), May 2019, pp. 1–6, doi: 10.1109/CFEC.2019.8733144.
[15] S. Ju et al., “Optimal county-level crop yield prediction using MODIS-based variables and weather data: A comparative study on
machine learning models,” Agricultural and Forest Meteorology, vol. 307, Sep. 2021, doi: 10.1016/j.agrformet.2021.108530.
[16] K. Tadesse et al., “Effects of preceding crops and nitrogen fertilizer on the productivity and quality of malting barley in tropical
environment,” Heliyon, vol. 7, no. 5, May 2021, doi: 10.1016/j.heliyon.2021.e07093.
[17] B. Kuzman et al., “Estimation of optimal fertilizers for optimal crop yield by adaptive neuro fuzzy logic,” Rhizosphere, vol. 18,
Jun. 2021, doi: 10.1016/j.rhisph.2021.100358.
[18] D. Elavarasan and P. M. D. Vincent, “Crop yield prediction using deep reinforcement learning model for sustainable Agrarian
applications,” IEEE Access, vol. 8, pp. 86886–86901, 2020, doi: 10.1109/ACCESS.2020.2992480.
[19] V. Tomar et al., “Rice equivalent crop yield assessment using MODIS sensors’ based MOD13A1-NDVI data,” IEEE Sensors
Journal, vol. 14, no. 10, pp. 3599–3605, Oct. 2014, doi: 10.1109/JSEN.2014.2329185.
[20] W.-J. Hu, J. Fan, Y.-X. Du, B.-S. Li, N. Xiong, and E. Bekkering, “MDFC-ResNet: An agricultural IoT system to accurately
recognize crop diseases,” IEEE Access, vol. 8, pp. 115287–115298, 2020, doi: 10.1109/ACCESS.2020.3001237.
[21] L. Momesso, C. A. C. Crusciol, H. Cantarella, K. S. Tanaka, G. A. Kowalchuk, and E. E. Kuramae, “Optimizing cover crop and
fertilizer timing for high maize yield and nitrogen cycle control,” Geoderma, vol. 405, Jan. 2022, doi:
10.1016/j.geoderma.2021.115423.
[22] A. Shan et al., “Effects of straw return with N fertilizer reduction on crop yield, plant diseases and pests and potential heavy metal
risk in a Chinese rice paddy: A field study of 2 consecutive wheat-rice cycles,” Environmental Pollution, vol. 288, Nov. 2021,
doi: 10.1016/j.envpol.2021.117741.
[23] S. Yang et al., “Integration of crop growth model and random forest for winter wheat yield estimation from UAV hyperspectral
imagery,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 14, pp. 6253–6269, 2021,
doi: 10.1109/JSTARS.2021.3089203.

6537
[24] P. Dayananda, M. Srikantaswamy, S. Nagaraju, R. Velluri, and D. M. Kumar, “Efficient detection of faults and false data
injection attacks in smart grid using a reconfigurable Kalman filter,” International Journal of Power Electronics and Drive
Systems (IJPEDS), vol. 13, no. 4, pp. 2086–2097, Dec. 2022, doi: 10.11591/ijpeds.v13.i4.pp2086-2097.
[25] S. Mallikarjunaswamy, “Implementation of an effective hybrid model for islanded microgrid energy management,” Indian
Journal of Science and Technology, vol. 13, no. 27, pp. 2733–2746, Jul. 2020, doi: 10.17485/IJST/v13i27.982.
[26] P. Satish, M. Srikantaswamy, and N. K. Ramaswamy, “A comprehensive review of blind deconvolution techniques for image
deblurring,” Traitement du Signal, vol. 37, no. 3, 2020.
[27] S. Thazeen, S. Mallikarjunaswamy, M. N. Saqhib, and S. N, “DOA method with reduced bias and side lobe suppression,” in 2022
International Conference on Communication, Computing and Internet of Things (IC3IoT), Mar. 2022, pp. 1–6, doi:
10.1109/IC3IOT53935.2022.9767996.
[28] S. Pooja, S. Mallikarjunaswamy, and N. Sharmila, “Adaptive sparsity through hybrid regularization for effective image
deblurring,” Indian Journal of Science and Technology, vol. 14, no. 24, pp. 2051–2068, Jun. 2021, doi:
10.17485/IJST/v14i24.604.
[29] S. Rathod, N. K. Ramaswamy, M. Srikantaswamy, and R. K. Ramaswamy, “An efficient reconfigurable peak cancellation model
for peak to average power ratio reduction in orthogonal frequency division multiplexing communication system,” International
Journal of Electrical and Computer Engineering (IJECE), vol. 12, no. 6, pp. 6239–6247, Dec. 2022, doi:
10.11591/ijece.v12i6.pp6239-6247.
[30] S. Thazeen and S. Mallikarjunaswamy, “The effectiveness of 6T beamformer algorithm in smart antenna systems for convergence
analysis,” IIUM Engineering Journal, vol. 24, no. 2, pp. 100–116, Jul. 2023, doi: 10.31436/iiumej.v24i2.2730.
[31] H. N. Mahendra and S. Mallikarjunaswamy, “An efficient classification of hyperspectral remotely sensed data using support
vector machine,” International Journal of Electronics and Telecommunications, vol. 68, 2022.
[32] H. N. Mahendra et al., “Deep learning models for inventory of agriculture crops and yield production using satellite images,” in
2022 IEEE 2nd Mysore Sub Section International Conference (MysuruCon), Oct. 2022, pp. 1–7, doi:
10.1109/MysuruCon55714.2022.9972523.
[33] H. N. Mahendra, S. Mallikarjunaswamy, C. B. Nooli, M. Hrishikesh, N. Kruthik, and H. M. Vakkalanka, “Cloud based
centralized smart cart and contactless billing system,” in 2022 7th International Conference on Communication and Electronics
Systems (ICCES), Jun. 2022, pp. 820–826, doi: 10.1109/ICCES54183.2022.9835856.
[34] S. Mallikarjunaswamy, N. Sharmila, G. K. Siddesh, K. R. Nataraj, and M. Komala, “A novel architecture for cluster based false
data injection attack detection and location identification in smart grid,” in Lecture Notes in Mechanical Engineering, Springer
Singapore, 2022, pp. 599–611.
BIOGRAPHIES OF AUTHORS
Nandini Gowda Puttaswamy currently working as Asst. Professor in
Information Science and Engineering, Sapthagiri College of Engineering, Bangalore. She
completed her B E in CSE from VTU, M. Tech. in Software Engineering from VTU and
pursuing Ph.D. from Visvesvaraya Technological University. She has published around 4
papers on National Conference and her area of research interest are cloud computing, fog
computing, edge computing and IoT, artificial intelligence (AI), machine learning (ML), big
data analytics. She can be contact at email: nandini.educator@gmail.com.
Anitha Narasimha Murthy currently working as Professor in Computer Science
and Engineering, BNM Institute of Technology, Bangalore. She completed her B E in CSE
from Bangalore University, M. Tech. in Information Technology from Bangalore University
and Ph.D. from Visvesvaraya Technological University. She has published around 30 research
papers and her area of research interest are AI, ML, big data analytics, data mining. She can be
contacted at email: anitha.mhp@gmail.com.

An efficient reconfigurable workload balancing scheme for fog computing network using internet of things devices

More Related Content

Similar to An efficient reconfigurable workload balancing scheme for fog computing network using internet of things devices (20)

More from IJECEIAES (20)

Recently uploaded (20)

An efficient reconfigurable workload balancing scheme for fog computing network using internet of things devices