IAES International Journal of Artificial Intelligence (IJ-AI)

Vol. 13, No. 4, December 2024, pp. 3849~3857

ISSN: 2252-8938, DOI: 10.11591/ijai.v13.i4.pp3849-3857  3849

A multi-core makespan model for parallel scientific workflow

execution in cloud computational framework

Farha Naaz, Sameena Banu

Department of Computer Science and Engineering, Khaja Banda Nawaz University, Kalaburagi, India

Article Info ABSTRACT

Article history: Researchers have shown a lot of potential in optimizing cloud-based
workload-scheduling over the past few years. However, executing scientific
Received Dec 28, 2023 workloads inside the cloud is time-consuming and costly, making it
Revised Mar 29, 2024 inefficient from both a financial and productivity standpoint. As a result,
Accepted Apr 17, 2024 there are many investigations conducted, with the general trend being to
speed up the rate of processing and establish a cost-effective system,
whereby customers are billed according to their actual use. In addition,
Keywords: energy-consumption is capable of being reduced, especially if the available
resources are heterogeneous; however, few investigations have optimized
Cloud computing multi-core with analyzing makespan parameters collectively to fulfill the
Energy quality of service (QoS) and service level agreement (SLA) of the workload
MapReduce task. In this research, we introduce an optimal scheduling for a
Quality of services heterogeneous distributed cloud computing environment called task aware
Scientific workflow makespan optimized scheduler (TAMOS) that guarantees requirements
across the task levels of scientific workflows. The energy and time required
to carry out specific workflows are significantly reduced by using this
TAMOS strategy. The TAMOS framework was studied using the scientific
workflows namely, inspiral and sipht. When compared to the conventional
method of scheduling work, our methodology used less energy and
makespan.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Farha Naaz
Department of Computer Science and Engineering, KBN University
Kalaburagi, India
Email: [email protected]

1. INTRODUCTION
Parallel scientific workflow tasks are extensively employed in both scientific and business
investigations [1]. Several instances of workloads in various scientific fields can be observed. These involve
astrophysics workloads such as Montage and Ligo, which focus on astronomical data analysis. Additionally,
there is a tremor identification workload called CyberShake, which aims to identify seismic activities.
Furthermore, deoxyribonucleic acid (DNA) sequencing workloads like epigenomics and sipht are employed
for studying genetic information. These workloads have been identified in previous research [2]. Multiple
scientific workloads are commonly utilized for a multitude of reasons. The execution of scientifically
challenging workloads is observed across various platforms, including MapReduce [3] and Amazon EC2 [4].
Platforms like MapReduce [5] as shown in Figure 1, are not efficient for execution of complex workflow
having multiple levels of dependency.
On the other side, the cloud-based parallel computational infrastructures as shown in Figure 2 are
known for their ability to offer top-notch storing and computing capabilities, including applications,
networks, and services. These resources are particularly valuable for handling the demanding computational

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

3850  ISSN: 2252-8938

requirements of scientific demanding workloads [6]. In contemporary times, various scientific disciplines,
such as science, biology, and astrophysics, have been increasingly leveraging cloud services to simulate and
analyze sophisticated scientific-workloads. This utilization of cloud-based parallel computational systems
enables researchers to devise better approaches for real-time challenges [7]. Therefore, it is evident that
scientific complicated tasks are effectively performed on cloud-based parallel computational systems,
wherein the evaluation is conducted utilizing computational tools offered by the cloud infrastructure.

Figure 1. Standard hadoop based computational platform

Figure 2. Cloud-based computational platform for execution of scientific workflow with multi-level
dependencies

Moreover, it has been observed that the implementation of scientifically complicated tasks in cloud
environments leads to a significant reduction in both execution cost and energy, as highlighted in previous
research [7]. The intricate nature of tasks can be effectively characterized by employing a directed acyclic
graph (DAG) framework. Within this framework, the centers of the DAG represent the interdependencies
among tasks, while the edges of the DAG delineate the specific tasks themselves. Through the
implementation of this process, it is anticipated that the workload can be effectively and efficiently carried

Int J Artif Intell, Vol. 13, No. 4, December 2024: 3849-3857

Int J Artif Intell ISSN: 2252-8938  3851

out within the designated deadlines. This implies that the tasks are going to be carried out in a sequential
manner, adhering to their predetermined order [8]. Furthermore, the intricate interdependencies among the
different tasks pose a significant challenge when it comes to effectively managing and allocating the
workload within a cloud computing environment.
In recent years, more and more people have started using cloud services for their various software
needs. As a result, a significant number of researchers have opted to utilize cloud services as a means to
efficiently manage and allocate their workload [9]. The provided diagram, Figure 2, illustrates a
straightforward architectural representation of the workload-scheduling process within a cloud environment.
Designing an effective workload-scheduling approach necessitates a thorough examination of current
systems, a process that in turn poses multiple difficulties. Notably, carrying out of more complicated and
bigger scientific-workloads demands increased execution-time and incurs higher costs. The execution of
tasks within a specified deadline presents an increased level of difficulty. Numerous investigations were
conducted by scholars from various institutions, focusing on the development of algorithms based on
heuristics. These algorithms aim to offer optimal solutions for a wide range of issues. Moreover, it has been
observed that the aforementioned heuristic algorithms exhibit a lack of time efficiency. Consequently, a
significant number of scholars have encountered challenges in attaining the best possible outcome, resulting
in adverse impacts on service level agreement (SLA) compliance and quality of service (QoS). Furthermore,
it has been established in the literature that the allocation of workload is classified as an NP-hard
(non-deterministic polynomial-time) problem [10]. Indeed, the optimization of both cost and time presents a
formidable challenge within the domain of workload-scheduling [11]. In the context of a scheduling
approach, it is observed that when the objective is to minimize costs, there is an associated rise in the time
necessary for finishing a specific task. The interdependence of time and cost is a significant contributing
factor to the observed phenomenon. The lack of consideration for virtual machine (VM) decision-making
procedures in current approaches during schedule creation has resulted in an ongoing issue of make-span and
cost [12], [13].
To effectively tackle the issue, this research introduces an innovative approach known as task aware
makespan optimized scheduler (TAMOS) under multi-core platform, which aims to ensure SLA compliance
at the task levels. The TMOS is designed in such a way it reduces makespan leveraging multi-core execution
and thereby utilizing system resource i.e., reducing overall energy consumption meeting multi-level scientific
workflow task dependencies. The significance of the research work is mentioned here.
− The work introduces a novel makespan model leveraging multi-core resource utilization.
− A makespan with task dependencies model considering multi-level workflow task interdependence of
scientific workflow,
− Simulation is considered using two different scientific workflows namely inspiral and sipht.
− Experiment outcome shows the proposed model exhibits superior performance in terms of reducing
makespan and energy in comparison with existing methodology.
The manuscript organization: section 2 introduces current state-of-the-art system to meets strict QoS
and SLA constraint of scientific workflow application with multi-level dependencies. Section 3 introduces a
novel makespan model to fully utilize resources efficiently and assures task dependencies of scientific
workflow application. The performance is proposed and existing is studied using two scientific workflows in
section 4. the last section the significance of research and future enhancement is provided.

2. LITERATURE SURVEY
The scientific workflow applications have grown exponentially in recent years due to availability of
parallel and distributed computing platforms, and their efficiency has improved as a result. Methods created
in [14] for a homogeneous framework execute unsatisfactorily in a heterogeneous framework due to the need
for the input/output (I/O) and memory optimization method, despite providing lockless first in first out
(FIFO) which incorporates hadoop map reduce (HMR) and other applications. In the HMR system, the
scheduler technique used during the shuffling stage is the primary determinant of makespan for completing
tasks [15]. As demonstrated in [16], prior approaches to performing heterogeneous tasks have not taken task
dependence models into account, leading to inefficient use of available resources [17]. So, they devised a
new yet another resource negotiator (YARN) scheduling scheme that cuts down the makespan of various
industrial tasks. Makespan efficiency, particularly when dealing with complicated repetitive applications, is
negatively impacted by the fact that the models described in [18], [19] don't account for failures at
intermediary tasks considering DAG applications. The optimization of energy and cost for diverse
computational structures is the main objective of the study [20]. The researchers aimed to develop a
scheduling mechanism that effectively managed the workload in order to achieve optimal outcomes in terms
of both energy and cost efficiency. The utilization of a minimum functionality in this context aims to

A multi-core makespan model for parallel scientific workflow execution in cloud … (Farha Naaz)
3852  ISSN: 2252-8938

optimize the use of energy and meet task-deadlines, taking into account the dispersed location. Distribution
of task-related data. The deadlines have been carefully examined and organized in ascending order, starting
from the smallest to the largest. In this study, they presented a novel approach to address the challenge of
selecting an effective schedule for execution of processes. The proposed approach was based on adaptable
searching techniques, which have shown promise in various optimization problems. By leveraging the
adaptability of the search algorithm, their method aimed to identify the most suitable schedule that can
maximize the efficiency of the method of execution.
This research contribution fills a gap in the existing literature by offering a practical solution to the
problem at hand. As per [12], there exists an immediate connection between escalated computation expenses
associated with service provision and the corresponding surge in energy usage. Timeliness and reliability are
widely recognized as the primary metrics of utmost significance within the realm of customer service supply.
The researchers devised a scheduling structure called energy min scheduling (EMS) to address the need for
reduced energy-consumption in running workloads. This framework successfully fulfilled both the timeliness
and reliability requirements. The application of non-linear mixed-integer-programming (NL-MIP) was
employed in this particular scenario to obtain an optimal solution. The present study focused on the
development of a meeting reliability approach, which centered on the utilization of a schedule length
minimizing approach. Furthermore, the implementation of dynamic-voltage frequency-scaling (DVFS)
method allowed for the reduction of energy-consumption through the method of processor merging with one
another In this particular case, scaling was performed at both the processor and task stages. The findings of
this study suggested that previous frameworks performed more effectively when exposed to varying degrees
of disruption.
Nested particle swarm optimization (NPSO) and fast nested particle swarm optimization (FNPSO)
are two evolutionary computing models developed in [21] to enhance the performance of demanding
workloads. Compared to the traditional NPSO approach, the FNPSO is considerably reduced make-span and
cost. As per the findings of the study conducted by [22], it is imperative for a cloud-based scheduling method
to fulfill both user time constraints and SLAs. A multi-cloud system [23] was employed in order to align with
the particular efficiency and cost criteria of the stream workload implementation. By leveraging a multi-
cloud structure and implementing a fault-tolerant scheduling architecture, a system can be developed to
address the task at hand. The utilization of a multi-cloud structure allows for the distribution of workloads
across multiple cloud service providers, thereby enhancing resource availability and scalability. Additionally,
the incorporation of a fault-tolerant scheduling structure ensures that the system can withstand and recover
from potential failures or disruptions, thereby minimizing downtime and maximizing reliability [24].
Furthermore, the proposed model not only guarantees the fulfillment of the reliability criterion but also
effectively reduces the associated cost. A comprehensive examination was conducted to assess the rates of
failure and reliability by employing an ongoing probability distribution. After determining the financial
implications of implementing the multi-cloud platform, the subsequent phase involves formulating a fault-
tolerant workload-scheduling framework that guarantees dependability, cost reduction, minimized execution
duration, and cost efficiency. The inability to fulfill the cost constraints of the application can be attributed to
the absence of load balancing mechanisms [25]. The subsequent section presents a scheduling strategy that
takes into consideration QoS performance factors like makespan, resource utilization and SLA constraints
like workflow deadline at different level in heterogeneous cloud environments, in order to tackle the
challenges previously mentioned.

3. PROPOSED METHODOLOGY
This paper introduces a novel cloud-based parallel computational framework for effective execution
of parallel scientific workflow with reduced makespan as shown in Figure 3. This work introduces a
multi-level workflow application specific QoS dependency aware makespan model. Namely task-aware
makespan optimized scheduler (TAMOS) meeting SLA constraint with better resource utilization leveraging
multi-core platform.
A comprehensive makespan approach is presented to address the issue of improving task execution
efficiency in the TAMOS framework. The computation of the makespan 𝐶 for carrying out a task is easily
achieved by utilizing the equation shown here. The computation of the makespan is mentioned in (1).

𝐶 = 𝐶𝑇 + 𝐶𝑀 + 𝐶𝑅 . (1)

The variable 𝐶𝑇 represents the makespan for the beginning worker considering both I/O and memory
optimization, while 𝐶𝑀 denotes the makespan for carrying out of map tasks, and 𝐶𝑅 represents the makespan

Int J Artif Intell, Vol. 13, No. 4, December 2024: 3849-3857

Int J Artif Intell ISSN: 2252-8938  3853

for carrying out reduce tasks. Assume a scenario where each worker, denoted as 𝑞, is comprised of a
particular number of cores/threads, represented by 𝑛, and has a memory dimension of 𝑥.

Figure 3. Architecture of proposed cloud-based computational framework for execution of parallel

scientific workflow

In this context, the mean make-span for carrying out a task is able to be determined through in (2).
In an identical manner, the computation of the mean make-span for the reduce task can be determined using
(3). The computation of the overall make-span of TAMOS can be achieved by utilizing (2) and (3) and it is
presented in (4).
𝑞
∑𝑎=1 𝐶𝑎_𝑀
𝐶𝑀 = (2)
𝑞

𝑞
∑𝑎=1 𝐶𝑎𝑅
𝐶𝑅 = . (3)
𝑞

𝑞
∑𝑎=1 (𝐶𝑎_𝑀 +𝐶𝑎_𝑅)
𝐶 = 𝐶𝑇 + . (4)
𝑞

The task execution of various phases within the TAMOS framework exhibits variability. Therefore,
we have established both upper limits and lower limits for the make-span required to execute task 𝐾. In (5)
used to calculate the overall make-span for the 𝐾 𝑡ℎ tasks within the TAMOS framework under the ideal case
scenario as mentioned here.
𝐿 𝐿 𝐿 𝑈 𝐿
𝑈 𝐾𝑙𝑖𝑚 = 𝑈𝑀𝑙𝑖𝑚 + 𝑈 𝑅𝑙𝑖𝑚 − (𝑈𝑀𝑙𝑖𝑚 − 𝑈 𝑀𝑙𝑖𝑚 ) (5)

𝐿
The variable 𝑈 𝐾𝑙𝑖𝑚 is used to denote the shortest make-span time required for the execution of task 𝐾.
𝐿 𝐿
Similarly, 𝑈𝑀𝑙𝑖𝑚 represents the shortest amount of time required for carrying out a map task, while 𝑈 𝑅𝑙𝑖𝑚 and
𝐿 𝑈
𝑈𝑀𝑙𝑖𝑚 indicate the bare essential time required for carrying out a reduce-task. On the other hand, 𝑈𝑀𝑙𝑖𝑚
represents the longest amount of time required for carrying out a map-task. In (5) is written in simplified
representation as mentioned in (6). In the TAMOS framework, the overall makespan of the 𝐾 𝑡ℎ job under the
most severe circumstances is determined by applying the subsequent equation as given in (7). The above
equation is written in simplified manner as mentioned in (8). In (9) provided below is utilized to derive the

A multi-core makespan model for parallel scientific workflow execution in cloud … (Farha Naaz)
3854  ISSN: 2252-8938

overall makespan of task 𝐾 using the TAMOS framework. The estimation of the overall make-span 𝑈 ⃗ 𝐾 for
carrying out task 𝐾 is calculated utilizing (6) and (8) as follows in (10). In (10) is simplified as mentioned in
(11). The final makespan considering multi-level workflow execution is obtained by modifying (11) is
mathematically given as mentioned in (12).
𝐿 𝐿 𝐿 𝑈 𝐿
𝑈 𝐾𝑙𝑖𝑚 = 𝑈𝑀𝑙𝑖𝑚 + 𝑈 𝑅𝑙𝑖𝑚 − (𝑈𝑀𝑙𝑖𝑚 − 𝑈 𝑀𝑙𝑖𝑚 ) (6)

𝑈 𝑈 𝑈 𝑈 𝐿
𝑈 𝐾𝑙𝑖𝑚 = 𝑈𝑀𝑙𝑖𝑚 + 𝑈 𝑅𝑙𝑖𝑚 − (𝑈𝑀𝑙𝑖𝑚 − 𝑈𝑀𝑙𝑖𝑚 ) (7)

𝑈 𝐿 𝑈
𝑈 𝐾𝑙𝑖𝑚 = 𝑈𝑀𝑙𝑖𝑚 + 𝑈 𝑅𝑙𝑖𝑚 (8)
𝑈 𝐿
(𝑈 𝐾𝑙𝑖𝑚 +𝑈𝐾𝑙𝑖𝑚 )
⃗𝐾 =
𝑈 (9)
2

𝑈 𝑈 𝑈 𝐿 𝐿 𝐿 𝑈 𝐿
((𝑈𝑀𝑙𝑖𝑚 +𝑈 𝑅𝑙𝑖𝑚 −(𝑈𝑀𝑙𝑖𝑚 − 𝑈𝑀𝑙𝑖𝑚 ))+(𝑈𝑀𝑙𝑖𝑚 +𝑈 𝑅𝑙𝑖𝑚 −(𝑈𝑀𝑙𝑖𝑚 − 𝑈 𝑀𝑙𝑖𝑚 )))
⃗𝐾 =
𝑈 (10)
2

𝐿 𝑈 𝐿 𝐿 𝑈
((𝑈𝑀𝑙𝑖𝑚 + 𝑈 𝑅𝑙𝑖𝑚 )+(2𝑈𝑀𝑙𝑖𝑚 +𝑈 𝑅𝑙𝑖𝑚 −𝑈𝑀𝑙𝑖𝑚 ))
⃗𝐾 =
𝑈 (11)
2

𝐿 𝐿 𝑈 𝑈
(3𝑈𝑀𝑙𝑖𝑚 +𝑈 𝑅𝑙𝑖𝑚 +𝑈 𝑅𝑙𝑖𝑚 −𝑈𝑀𝑙𝑖𝑚 )
⃗𝐾 =
𝑈 (12)
2

The present study employs a methodology comparable to that described in previous works,
specifically [3], [5]. To construct a framework for data dependence through the utilization of regression
analysis. The TAMOS framework has been observed to exhibit superior performance in terms of minimizing
makespan and reducing energy by utilizing resource efficiently for the execution of parallel scientific
workflow which is shown through simulation study in next section.

4. RESULT AND DISCUSSION

The following section focuses on the evaluation of performance metrics, namely makespan, and
energy efficiency, in the context of the suggested TAMOS framework compared to the existing reliable
workflow scheduling (RWS) approach [13]. TAMOS and RWS have been developed and deployed through
the utilization of the Java programming language. The CloudSim Software-defined networking-based
network function virtualization (CloudSimSDN‐NFV) simulator [26] has been employed as the underlying
platform for the implementation of these systems. The inspiral and sipht scientific workflow is used for
verifying proposed TAMOS framework. The mathematical complexity of the tasks at hand involves a
significant amount of memory, processor, and input/output operations. The inspiral structure is known for its
computationally demanding nature, particularly with regards to central processing unit (CPU) and memory
usage. On the other side, the sipht structure is known for its computationally demanding nature, particularly
with regards to CPU and I/O usage.

4.1. Makespan performance

The following section addresses the analysis regarding the makespan required to complete the
execution of inspiral workflows with size ranging from 30, 50, 100, and 1000. Alongside, the section
addresses the analysis regarding the makespan required to complete the execution of sipht workflows with
size ranging from 30, 60, 100, and 1000. Figure 4 illustrates a visual representation of the makespan achieved
over the execution of the inspiral workflow employing the TAMOS and RWS scheduling algorithms, while
considering a diverse range of workflow scenarios such as 30, 50, 100. Figure 5 illustrates a visual
representation of the makespan achieved over the execution of the sipht workflow employing the TAMOS
and RWS scheduling algorithms, while considering a diverse range of workflow scenarios such as 30, 60,
100. Similarly, for inspiral workflow size of 1000 the visual representation of makespan is given in Figure 6
and for sipht workflow size of 1000 the visual representation of makespan is given in Figure 7. The
utilization of TAMOS, as compared to RWS, results in a notable enhancement in average makespan
efficiency, with an observed increase of 83.32% and 51.46% for inspiral and sipht workflow, respectively.

Int J Artif Intell, Vol. 13, No. 4, December 2024: 3849-3857

Int J Artif Intell ISSN: 2252-8938  3855

MAKESPAN-INSPIRAL MAKESPAN-SIPHT
MAKESPAN (SECONDS)

MAKESPAN (SECONDS)
RWS TAMOS RWS TAMOS

14213.24

11712.58
7870.87
7941.77

5016.97
4678.92
4449.16

3746.13
4469.22

1540.54
1425.96
1350.14

30 50 100 30 60 100
WORKFLOW SIZE WORKFLOW SIZE

Figure 4. Makespan efficiency with different Figure 5. Makespan efficiency with different sipht
inspiral workflow size workflow size

MAKESPAN-INSPIRAL VERY LARGE MAKESPAN-SIPHT VERY LARGE

MAKESPAN (SECONDS)
MAKESPAN (SECONDS)

117255.34
153822.04

RWS RWS
TAMOS TAMOS
11860.26

8989.29
1000 1000
WORKFLOW SIZE WORKFLOW SIZE

Figure 6. Makespan efficiency with inspiral Figure 7. Makespan efficiency with sipht workflow
workflow size of 1000 size of 1000

4.2. Energy usage

The following section addresses the analysis regarding the energy needed to complete the execution of
inspiral workflows with size ranging from 30, 50, 100, and 1000. Alongside, the work addresses the analysis
regarding the energy needed to complete the execution of sipht workflows with size ranging from 30, 50, 100,
and 1000. Figure 8 illustrates a visual representation of the energy efficiency achieved over the execution of the
inspiral workflow employing the TAMOS and RWS scheduling algorithms, while considering a diverse range
of workflow scenarios such as 30, 50, 100. Figure 9 illustrates a visual representation of the energy efficiency
achieved over the execution of the sipht workflow employing the TAMOS and RWS scheduling algorithms,
while considering a diverse range of workflow scenarios such as 30, 60, 100. Similarly, for inspiral workflow
size of 1000 the visual representation of energy usage is given in Figure 10 and for sipht workflow size of 1000
the visual representation of energy usage is given in Figure 11. The resource utilization by reducing energy
consumption of TAMOS, as compared to RWS, results in a notable enhancement in average energy efficiency,
with an observed increase of 92.64% and 72.87% for inspiral and sipht workflow, respectively.

ENERGY CONSUMPTION-INSPIRAL ENERGY CONSUMPTION-SIPHT

ENERGY CONSUMED (WATT)
ENERGY CONSUMPTION (WATT)

RWS TAMOS
20859.04

RWS TAMOS
41822.38

11389.74
16857.2

4376.65

3179.26
3159.21
3518.37

2820.99
1493.82
829.35
704.18

30 50 100 30 60 100
WORKFLOW SIZE WORKFLOW SIZE

Figure 8. Energy efficiency with different inspiral Figure 9. Energy efficiency with different sipht
workflow size workflow size

A multi-core makespan model for parallel scientific workflow execution in cloud … (Farha Naaz)
3856  ISSN: 2252-8938

ENERGY CONSUMPTION- ENERGY CONSUMPTION-SIPHT

INSPIRAL VERY LARGE VERY LARGE
ENERGY CONSUMPTION (WATT)

ENERGY CONSUMED (WATT)

RWS

1071246.93
1499121.69
RWS
TAMOS TAMOS

13653.451

11291.22
1000 1000
WORKFLOW SIZE WORKFLOW SIZE

Figure 10. Energy efficiency with inspiral Figure 11. Energy efficiency with sipht workflow
workflow size of 1000 size of 1000

5. CONCLUSION
The present study introduces a novel multi-core resource utilization aware makespan model for
workflow scheduling which successfully guarantees the fulfillment of task-level SLAs. To our knowledge,
very limited work has ever before addressed task-level SLA when planning a workflow scheduling. The
present study presents an innovative approach to adaptable scheduling, which aims to minimize total
makespan while simultaneously ensuring highest resource efficiency by reducing overall energy by providing
execution at thread-level in multi-core processor of cloud VM. By implementing this method, a substantial
reduction in overall operation makespan and energy thereby will aid in utilizing resources efficiently with
minimal cost can be achieved. TAMOS demonstrates high efficiency in the allocation of CPU, memory, and
I/O resources for executing parallel scientific workflows that require significant computational and data
processing capabilities. This efficiency is achieved by utilizing shared computing environments, particularly
in cloud environments. The experimental results demonstrate that the TAMOS system exhibits a high level of
makespan and energy efficiency for execution of cybershake workflows. Specifically, when compared to the
RWS system, TAMOS showcases an impressive increase of 67.39% and 82.75% for makespan and energy
efficiency, respectively. In the future, it is anticipated that the suggested scheduling approach will undergo
testing using a broader dataset of workloads. In addition, it is worth exploring the potential benefits of
utilizing an edge-cloud structure in order to potentially achieve cost reduction and minimize execution
delays.

REFERENCES
[1] J. Kim, H. Roh, and S. Park, “Selective I/O bypass and load balancing method for write-through SSD caching in big data
analytics,” IEEE Transactions on Computers, vol. 67, no. 4, pp. 589–595, Apr. 2018, doi: 10.1109/TC.2017.2771491.
[2] N. Zhang, M. Wang, Z. Duan, and C. Tian, “Verifying properties of mapreduce-based big data processing,” IEEE Transactions on
Reliability, vol. 71, no. 1, pp. 321–338, Mar. 2022, doi: 10.1109/TR.2020.2999441.
[3] D. Yang, D. Cheng, W. Rang, and Y. Wang, “Joint optimization of mapreduce scheduling and network policy in hierarchical data
centers,” IEEE Transactions on Cloud Computing, vol. 10, no. 1, pp. 461–473, Jan. 2022, doi: 10.1109/TCC.2019.2961653.
[4] A. Kumar, N. Varshney, S. Bhatiya, and K. U. Singh, “Replication-based query management for resource allocation using hadoop
and MapReduce over big data,” Big Data Mining and Analytics, vol. 6, no. 4, pp. 465–477, Dec. 2023, doi:
10.26599/BDMA.2022.9020026.
[5] X. Li, F. Chen, R. Ruiz, and J. Zhu, “MapReduce task scheduling in heterogeneous geo-distributed data centers,” IEEE
Transactions on Services Computing, vol. 15, no. 6, pp. 3317–3329, Nov. 2022, doi: 10.1109/TSC.2021.3092563.
[6] R. Jeyaraj and A. Paul, “Optimizing MapReduce task scheduling on virtualized heterogeneous environments using ant colony
optimization,” IEEE Access, vol. 10, pp. 55842–55855, 2022, doi: 10.1109/ACCESS.2022.3176729.
[7] X. Li, W. Yu, R. Ruiz, and J. Zhu, “Energy-aware cloud workflow applications scheduling with geo-distributed data,” IEEE
Transactions on Services Computing, vol. 15, no. 2, pp. 891–903, Mar. 2022, doi: 10.1109/TSC.2020.2965106.
[8] P. Pujar, A. Kumar, and V. Kumar, “Efficient plant leaf detection through machine learning approach based on corn leaf image
classification,” IAES International Journal of Artificial Intelligence (IJ-AI), vol. 13, no. 1, pp. 1139–1148, Mar. 2024, doi:
10.11591/ijai.v13.i1.pp1139-1148.
[9] S. H. Sreedhara, V. Kumar, and S. Salma, “Efficient big data clustering using adhoc fuzzy C means and auto-encoder CNN,” in
Inventive Computation and Information Technologies: Proceedings of ICICIT 2022, Singapore: Springer, 2023, pp. 353–368, doi:
10.1007/978-981-19-7402-1_25.
[10] S. Qin, D. Pi, Z. Shao, Y. Xu, and Y. Chen, “Reliability-aware multi-objective memetic algorithm for workflow scheduling
problem in multi-cloud system,” IEEE Transactions on Parallel and Distributed Systems, vol. 34, no. 4, pp. 1343–1361, Apr.
2023, doi: 10.1109/TPDS.2023.3245089.

Int J Artif Intell, Vol. 13, No. 4, December 2024: 3849-3857

Int J Artif Intell ISSN: 2252-8938  3857

[11] S. Qin, D. Pi, Z. Shao, and Y. Xu, “A knowledge-based adaptive discrete water wave optimization for solving cloud workflow
scheduling,” IEEE Transactions on Cloud Computing, vol. 11, no. 1, pp. 200–216, Jan. 2023, doi: 10.1109/TCC.2021.3087642.
[12] E. Khodayarseresht and A. S. -Sendi, “A multi-objective cloud energy optimizer algorithm for federated environments,” Journal
of Parallel and Distributed Computing, vol. 174, pp. 81–99, Apr. 2023, doi: 10.1016/j.jpdc.2022.12.007.
[13] X. Tang, “Reliability-aware cost-efficient scientific workflows scheduling strategy on multi-cloud systems,” IEEE Transactions
on Cloud Computing, vol. 10, no. 4, pp. 2909–2919, Oct. 2022, doi: 10.1109/TCC.2021.3057422.
[14] G. Papadimitriou et al., “End-to-end online performance data capture and analysis for scientific workflows,” Future Generation
Computer Systems, vol. 117, pp. 387–400, Apr. 2021, doi: 10.1016/j.future.2020.11.024.
[15] R. F. D. Silva, H. Casanova, A. -C. Orgerie, R. Tanaka, E. Deelman, and F. Suter, “Characterizing, modeling, and accurately
simulating power and energy consumption of I/O-intensive scientific workflows,” Journal of Computational Science, vol. 44, Jul.
2020, doi: 10.1016/j.jocs.2020.101157.
[16] S. Hedayati, N. Maleki, T. Olsson, F. Ahlgren, M. Seyednezhad, and K. Berahmand, “MapReduce scheduling algorithms in
Hadoop: a systematic study,” Journal of Cloud Computing, vol. 12, no. 1, Oct. 2023, doi: 10.1186/s13677-023-00520-9.
[17] N. Garg, Neeraj, M. Raj, I. Gupta, V. Kumar, and G. R. Sinha, “Energy-efficient scientific workflow scheduling algorithm in
cloud environment,” Wireless Communications and Mobile Computing, vol. 2022, pp. 1–12, Mar. 2022, doi:
10.1155/2022/1637614.
[18] J. Wang, X. Li, R. Ruiz, J. Yang, and D. Chu, “Energy utilization task scheduling for MapReduce in heterogeneous clusters,”
IEEE Transactions on Services Computing, vol. 15, no. 2, pp. 931–944, Mar. 2022, doi: 10.1109/TSC.2020.2966697.
[19] Q. Liu et al., “Cloud, edge, and mobile computing for smart cities,” in Urban Informatics, Singapore: Springer, 2021, pp. 757–
795, doi: 10.1007/978-981-15-8983-6_41.
[20 S. Danthuluri and S. Chitnis, “Energy and cost optimization mechanism for workflow scheduling in the cloud,” Materials Today:
Proceedings, vol. 80, pp. 3069–3074, 2023, doi: 10.1016/j.matpr.2021.07.168.
[21] N. Bacanin, M. Zivkovic, T. Bezdan, K. Venkatachalam, and M. Abouhawwash, “Modified firefly algorithm for workflow
scheduling in cloud-edge environment,” Neural Computing and Applications, vol. 34, no. 11, pp. 9043–9068, Jun. 2022, doi:
10.1007/s00521-022-06925-y.
[22] X. Wang, J. Cao, and R. Buyya, “Adaptive cloud bundle provisioning and multi-workflow scheduling via coalition reinforcement
learning,” IEEE Transactions on Computers, vol. 72, no. 4, pp. 1041–1054, Apr. 2023, doi: 10.1109/TC.2022.3191733.
[23] J. Anselmi and N. Walton, “Stability and optimization of speculative queueing networks,” IEEE/ACM Transactions on
Networking, vol. 30, no. 2, pp. 911–922, Apr. 2022, doi: 10.1109/TNET.2021.3128778.
[24] J. P. -Valero, A. Banchs, P. Serrano, J. Ortín, J. G. -Reinoso, and X. C. -Pérez, “Energy-aware adaptive scaling of server farms for
NFV with reliability requirements,” IEEE Transactions on Mobile Computing, vol. 23, no. 5, pp. 4273–4284, May 2024, doi:
10.1109/TMC.2023.3288604.
[25] B. Hu, Z. Cao, and M. Zhou, “Energy-minimized scheduling of real-time parallel workflows on heterogeneous distributed
computing systems,” IEEE Transactions on Services Computing, vol. 15, no. 5, pp. 2766–2779, Sep. 2022, doi:
10.1109/TSC.2021.3054754.
[26] J. Son, T. He, and R. Buyya, “CloudSimSDN‐NFV: Modeling and simulation of network function virtualization and service
function chaining in edge computing environments,” Software: Practice and Experience, vol. 49, no. 12, pp. 1748–1764, Dec.
2019, doi: 10.1002/spe.2755.

BIOGRAPHIES OF AUTHOR

Farha Naaz earned her B.E. degree in CSE from K.B.N College of Engineering
under VTU University, Belagavi in 2012 and master degree in M.Tech. in CSE from K.B.N
College of Engineering under VTU University in 2015. Currently she is research scholar at
K.B.N University doing her Ph.D. in CSE. Her areas of interest are big data analytics, cloud
computing, and computer networks. She can be contacted at email:
[email protected].

Sameena Banu is an associate professor in the Department of Computer Science

and Engineering, Faculty of Engineering and Technology, Khaja Banda Nawaz University,
Kalaburgi. She is qualified in bachelor and master degree in Computer Science and
Engineering and Ph.D. in Computer Science and Engineering in the area of digital image
processing. She is having 21 years of teaching experience. Her areas of interest are image
processing, machine learning, artificial intelligence, and internet of things. She can be
contacted at email: [email protected].

A multi-core makespan model for parallel scientific workflow execution in cloud … (Farha Naaz)