INDIAN JOURNAL OF SCIENCE AND TECHNOLOGY

RESEARCH ARTICLE

Intelligent Particle Swarm Optimization

Based Resource Provisioning Technique
in Cloud Computing
OPEN ACCESS B Ramana Reddy1,2 ∗ , M Indiramma3 , N Nagarathna3
Received: 12-02-2023 1 Research Scholar, Department of CSE, B.M.S. College of Engineering, Bangalore, India
Accepted: 04-04-2023 2 Assistant Professor, Department of CSE, Chaitanya Bharathi Institute of Technology,
Hyderabad, India
Published: 28-04-2023
3 Professor, Department of CSE, B.M.S. College of Engineering, Bangalore, India

Citation: Reddy BR, Indiramma M, Abstract

Nagarathna N (2023) Intelligent
Particle Swarm Optimization Based Objectives: To find the optimal allocation of resources and minimize the
Resource Provisioning Technique in overall cost while meeting the performance requirements of the applications.
Cloud Computing. Indian Journal of
Methods: The proposed Intelligent PSO-based resource optimization in cloud
Science and Technology 16(16):
1241-1249. https://doi.org/ computing evaluates the quality of the solutions based on their resource
10.17485/IJST/v16i16.308 allocation parameters. Cloud Sim software is used as a simulation tool
∗
Corresponding author. for testing and evaluating new solutions and strategies in the cloud. The
Closest Data Center Service Broker Policy is implemented in Cloud Simulation.
[email protected]
Findings: The proposed technique assigns workloads effectively on available
Funding: None resources with an improvement of 10.46% in electricity consumption. Novelty:
Competing Interests: None The algorithm can further be employed for identifying the unused VM’s in the
Copyright: © 2023 Reddy et al. This Data Center to reduce the cost.
is an open access article distributed Keywords: Cloud computing; Resource Scheduling; Load balancing; Particle
under the terms of the Creative
Swarm Optimization; Quality of Service
Commons Attribution License, which
permits unrestricted use,
distribution, and reproduction in
any medium, provided the original 1 Introduction
author and source are credited.
In the field of computer science, cloud computing technology is showing phenomenal
Published By Indian Society for growth due to the advancement of the internet (1) . Cloud computing provides
Education and Environment (iSee) infrastructure, platform, and software as services. Cloud resources are providing
ISSN customers with a pay-as-you-use model. To provide quality services to customers,
Print: 0974-6846 there is a Service Level Agreement (SLA) between the customers and the cloud service
Electronic: 0974-5645 providers as shown in Figure 1. Cloud service providers need to verify if a sufficient
number of resources are available to customers to ensure that QoS requirements like
execution time, deadline, and budget restrictions are met. However, running too many
applications on a single resource can result in a drastic performance loss, which deters
cloud consumers. It is challenging to match workloads to the right resources for
cloud execution. The existing approaches of Resource optimization in cloud computing
involves managing multiple resources and balancing different performance metrics,
which can be complex and challenging. Developing an effective optimization algorithm
requires significant expertise and resources. For effective resource use, three primary
QoS limitations must be taken into account (2) .

https://www.indjst.org/ 1241
Reddy et al. / Indian Journal of Science and Technology 2023;16(16):1241–1249

1. Must meet deadlines with minimum execution time and cost

2. Minimum electricity consumption
3. User satisfaction
A PSO-based resource scheduling algorithm, schedules workloads in a cloud environment to save execution time, cost, and
energy. By figuring out how workloads and resources interact in a cloud setting was expanded. Experimental findings enhance
QoS metrics like resource availability, reliability, latency, and resource consumption to the maximum extent possible (3) .
The Particle Swarm Optimization and Genetic Algorithm (PSO-GA) was used to make runtime decisions for exploring the
objectives of the resource allocation plan (4) . The issue of resource allocation for cloud-based services is addressed by introducing
the workload time window. Further, it builds a calculation model for optimizing VM resource allocation plans (5) . The PSO-GA-
based method observed the advantages of PSO and GA and improved their inadequacies in population diversity, search range,
and convergence speed. The algorithm strategy objective takes the current and future workloads into the process of producing
resource allocation plans.
In the proposed research work, the intelligent Particle Swarm Optimization focuses on minimizing execution time, cost and
electricity consumption.

Fig 1. Resource Request in Cloud Computing

2 Methodology
2.1 Proposed Intelligent Particle Swarm Optimization-based resource
2.1.1 Scheduling Technique
In cloud computing, resource scheduling is important for resource management (6) . It directs the allocation of cloud workloads
to cloud resources. After, applications are scheduled to cloud resources using an Intelligent Particle Swarm optimization-based
heuristic framework, which lowers the cost of computation and data transfer. Most existing research takes into account of cloud
computing’s fundamental characteristics in order to execute heterogeneous cloud work loads quickly and affordably (7) . The
population in the I-PSO algorithm is defined as particles, where the particles are initialized randomly. In every new generation,
the fitness value of each particle is calculated and the two values of the particles need to be calculated: 1. Local best(Lbest ), 2.
Global best(Gbest ). Where the Lbest of a particle is the best result reached by the particle so far, and the Gbest is the best result
among the whole population. Intelligent PSO optimization technique works on global search (8) . Every single particle controls
its own independent course according to the local and global best in every generation. Numerous NP-hard issues, including
task distribution and resource scheduling, will be resolved by intelligent PSO.
Workloads : Workloads that are initiated by the user are placed in a queue for execution and processing purposes.
Resource Manager:The resource manager maintains information about resources, QoS, and SLA.
Quality of Service: Quality of Services like availability, latency, resource utilization, and reliability must be maintained.
Service Level Agreement : Provides information about suitable service level agreements between customers and cloud
service providers.

https://www.indjst.org/ 1242
Reddy et al. / Indian Journal of Science and Technology 2023;16(16):1241–1249

Workload Analyzer: Workload Analyzer identifies different characteristics of cloud workloads.

Resource details: Maintains resource details such as availability of virtual machines, size of virtual memory, cost of cloud
resources, and type of cloud resource.
Resource provisioner: It provides workload to the demanded resources for their execution in the cloud only if the resources
are available in the resource pool.
Resource scheduler: It executes all the workloads on a provisioned resource efficiently.
The Proposed architecture of resource optimization is described in Figure 2.

Fig 2. Architecture of proposed Resource optimization

2.1.2 Objective Function

The fitness value is calculated based on execution time, cost, and electricity consumption as shown in equation 1.

fitness = c1 ∗ execution time + c2 ∗ execution cost + c3 ∗ electricitycon (1)

0 ≤ c1 ≤ 1, 0 ≤ c2 ≤ 1 and 0 ≤ c3 ≤ 1
where c1, c2, and c3 are weights to prioritize the components of the fitness function.
Execution time is the time taken to execute the workload on the allocated resources as follows:

Execution time = min (t (wi , ri )) for 1 ≤ i ≤ n (2)

https://www.indjst.org/ 1243
Reddy et al. / Indian Journal of Science and Technology 2023;16(16):1241–1249

where t(wi , ri ) is the time taken by workload wi executed by resource ri .

Execution cost is the cost of the workloads executed by the assigned resources in the following equation.

Execution cost = min (c (wi , ri )) for 1 ≤ i ≤ n (3)

where c(wi , ri ) is the cost of workload wi executed by resource ri . electricity consumption is the electricity consumed in resource
utilization calculated by

electricity co n = electricity vm + electricity memory + electricity misc (4)

Where electricityvm is the virtual machine electricity consumption, electricitymemory is the electricity consumption for memory
operations and electricitymisc is the electricity consumption for fans and other miscellaneous parts.
( )
electricity vm = vmidle + vmrunning − vmidle × vmprocessor (5)

where vmidle denotes the idle state of the virtual machine, vmrunning denotes the virtual machine running time power
consumption, and vm processor denotes the virtual machine processor capacity.

2.2 Proposed Algorithm

In this section, the description regarding I-PSO based algorithm for resource provisioning technique in the cloud environment
is shown in Algorithm 1.
Algorithm 1: Proposed Intelligent PSO-based resource scheduling Algorithm
Result: intelligent mapping of workloads to the VMs.
1 initialize VMs
2 initialize workloads
3 initialize a random feasible solution
4 for i=1 to PS
5 do
6 Pv ← RV
7 Pp ← RP
8 if fitness(Gbest ) > f itness(P )
9 then
10 Gbest ← P
11 end
12 end
13 while iterations are not reached max
14 do
15 for P ∈Ppop
16 do
17 Pv ← update velocity
18 Pp ← update position
19 if f itness(P ) < f itness(Lbest )
20 then
21 Lbest ← P
22 end
23 if f itness(Lbest) < f itness(Gbest)
24 then
25 Gbest ← Lbest
26 end
27 end
28 return(Gbest)
29 end
30 if any unassigned VMs then
31 VMs moved to sleep mode.

https://www.indjst.org/ 1244
Reddy et al. / Indian Journal of Science and Technology 2023;16(16):1241–1249

32 end
Intelligent PSO Terminology:
The notations used in an I-PSO algorithm are presented in Table 1.

• Particle: A particle is similar to a flock of birds searching for food. Every particle has velocity. Fitness values calculate a
particle’s performance. For the proposed I-PSO algorithm, cloud workloads are considered particles.
• Population size: In the proposed I-PSO, available resources in the cloud are considered as population size.
• Random velocity: Every particle’s velocity is updated with Lbest and Gbest values.
• Particle velocity:The particle’s velocity is calculated based on particle position.
• Particle position: positions depend on the submission status, waiting state, ready state, execution state, and completion
state.
• Global best (Gbest ): Best position of a particle among the whole group of particles.
• Local best (Lbest ): Best position reached by a particle

Table 1. Notations and its description

Notation Description
S Number of particles
PS Population size
RV Random velocity
RP Random position
Pv Particle velocity
Pp Particle position
Ppop Particle population
Gbest Global best position
Lbest Local best position

In Figure 3, all the algorithm steps are described in the form of a flow chart.

3 Results and Discussion

The experiment has been conducted in a simulator with size of 700MB image conversion cloud workload. Once the simulation
starts a sequence of steps will be execute as per the mentioned order.

1. Cloud user submits workload details like name and type of workload. Workload analyzer will submit these details to
Resource Manager.
2. Resource manager processes workload details and then asks the user for budget restrictions and deadline details.
3. Once the cloud user submits budget and deadline restrictions the Resource manager will generate tentative schedule and
cost.
4. Cloud user sends a confirmation of the details, SLA agreements to the Resource Manager (9) .
5. Cloud user pays requested amount and executes the workloads on allocated resources

The performance of the proposed I-PSO resource provisioning technique is compared with the existing scheduling
algorithms (10,11) .
Results – Execution time: In Table 2 results compared at 45 workloads, execution time in I-PSO is 2.74% lesser than ACO,
4.31% lesser than GA. At 90 workloads, execution time in I-PSO is 3.53% lesser than ACO, 5% lesser than GA. Figure 4 shows
that execution time of I-PSO is better than ACO and GA.
Results – Execution cost: In Table 3 results compared at 45 workloads, the cost incurred in I-PSO is $260, ACO is $266 and
GA is $292. At 90 workloads the cost incurred in I-PSO is $412, ACO is $422 and GA is $442. Figure 5 shows that the execution
cost of I-PSO is minimum compared to the ACO and GA.
Results – Electricity consumption: In Table 4 results compared at 45 workloads the I-PSO consumes 8.97% lesser than
ACO, 14.10% lesser than GA. At 90 workloads the I-PSO consumes 10.46% lesser than ACO, 16.27% lesser than GA. Figure 6
shows that the electricity consumption is minimum compared to the ACO and GA.

https://www.indjst.org/ 1245
Reddy et al. / Indian Journal of Science and Technology 2023;16(16):1241–1249

Fig 3. Flowchart of the proposed system

https://www.indjst.org/ 1246
Reddy et al. / Indian Journal of Science and Technology 2023;16(16):1241–1249

Table 2. Comparison of I-PSO Execution time (ms) with GA and ACO works
Execution Time (ms)
No. of work loads
GA ACO I-PSO
15 225 221 198
30 410 402 380
45 532 524 510
60 698 788 765
75 950 948 918
90 1218 1201 1160

Fig 4. Execution time vs no.of work loads

Table 3. Comparison of I-PSO Execution cost($) with GA and ACO works

Execution cost ( $)
No. of work loads
GA ACO I-PSO
15 196 176 169
30 230 204 198
45 292 266 260
60 332 307 297
75 368 346 336
90 442 422 412

Fig 5. Execution cost vs no.of work loads

https://www.indjst.org/ 1247
Reddy et al. / Indian Journal of Science and Technology 2023;16(16):1241–1249

Table 4. Comparison of I-PSO electricity consumption (kwh) with GA and ACO works
Electricity Consumption (Kwh)
No. of work loads
GA ACO I-PSO
15 82 76 70
30 88 82 76
45 89 85 78
60 92 87 81
75 96 92 84
90 100 95 86

Fig 6. Electricity consumption vs no.of workloads

4 Conclusion
In this research work, different techniques such as load balancing and resource allocation methods are provided which ensures
that resources are used optimally. Furthermore, the Proposed Intelligent PSO algorithm can help predict idle virtual machines,
enabling providers to proactively allocate resources. However, the I-PSO algorithm is unable to predict the upcoming workloads
due to ongoing changes of user requirements.
To further improve this research work, the computation of the fitness value can be more accurate by considering the priority
of virtual machines. Thus, the highest priority virtual machine will be assigned first and this result in a lower execution time. The
future scope will require a continuous improvement and innovation to address emerging challenges and technologies. One key
area of focus will be the development of several sophisticated machine learning algorithms that can accurately predict resource
demands and dynamically allocate resources in real-time. This will enable cloud providers to achieve higher levels of efficiency,
reduce costs, and improve overall performance.

References
1) Mishra SK, Sahoo B, Parida PP. Load balancing in cloud computing: A big picture. Journal of King Saud University - Computer and Information Sciences.
2020;32(2):149–158. Available from: https://doi.org/10.1016/j.jksuci.2018.01.003.
2) Singh A, Kumar R. Performance Evaluation of Load Balancing Algorithms Using Cloud Analyst. 2020 10th International Conference on Cloud Computing,
Data Science & Engineering (Confluence). 2020;p. 156–162. Available from: https://doi.org/10.1109/Confluence47617.2020.9058017.
3) Kruekaew B, Kimpan W. Multi-Objective Task Scheduling Optimization for Load Balancing in Cloud Computing Environment Using Hybrid Artificial Bee
Colony Algorithm With Reinforcement Learning. IEEE Access. 2022;10:17803–17818. Available from: https://doi.org/10.1109/ACCESS.2022.3149955.
4) Chen Z, Yang L, Huang Y, Chen X, Zheng X, Rong C. PSO-GA-Based Resource Allocation Strategy for Cloud-Based Software Services With Workload-
Time Windows. IEEE Access. 2020;8:151500–151510. Available from: https://doi.org/10.1109/ACCESS.2020.3017643.
5) Reddy RB, Indiramma M. Efficient Throttled load balancing algorithm to improve the response time and processing time in data center. Concurrency
And Computation-Practice & Experience. 2022;34(23). Available from: https://doi.org/10.1002/cpe.7208.
6) Malik N, Sardaraz M, Tahir M, Shah B, Ali G, Moreira F. Energy-Efficient Load Balancing Algorithm for Workflow Scheduling in Cloud Data Centers
Using Queuing and Thresholds. Applied Sciences. 2021;11(13):5849. Available from: https://doi.org/10.3390/app11135849.
7) Sweekriti M, Shetty S, Shetty. Analysis of load balancing in cloud data centers. Journal of Ambient Intelligence and Humanized Computing. 2019;p. 1–9.
Available from: https://doi.org/10.1007/s12652-018-1106-7.

https://www.indjst.org/ 1248
Reddy et al. / Indian Journal of Science and Technology 2023;16(16):1241–1249

8) Kothilaxman RR, Lathigara A, Aluvalu R, Viswanadhula UM. PGWO-AVS-RDA: An intelligent optimization and clustering based load balancing model
in cloud. 2022. Available from: https://doi.org/10.1002/cpe.7136.
9) Samriya JK, Patel SC, Khurana M, Tiwari PK, Cheikhrouhou OK. Intelligent SLA-Aware VM Allocation and Energy Minimization Approach with EPO
Algorithm for Cloud Computing Environment. Mathematical Problems in Engineering. 2021;2021:1–13. Available from: https://doi.org/10.1155/2021/
9949995.
10) Singh H, Tyagi S, Kumar P, Gill SS, Buyya R. Metaheuristics for scheduling of heterogeneous tasks in cloud computing environments: Analysis,
performance evaluation, and future directions. Simulation Modelling Practice and Theory. 2021;111:102353. Available from: https://doi.org/10.1016/j.
simpat.2021.102353.
11) Ragmani A, Elomri A, Abghour N, Moussaid K, Rida M. An improved Hybrid Fuzzy-Ant Colony Algorithm Applied to Load Balancing in Cloud
Computing Environment. Procedia Computer Science. 2019;151:519–526. Available from: https://doi.org/10.1016/j.procs.2019.04.070.

https://www.indjst.org/ 1249