Proceedings Volume of the 6th IET International Smart Cities Symposium,

6th SCS-2022, December 6-8, 2022, Bahrain

Effective Comparison of Meta-heuristics Algorithm in Cloud

Computing
Nihar Ranjan Nayak1, Subrata Chowdhury2, Srinivas Mishra3, Pushan Kumar Dutta4, Ramya G5

Department of CSE, Presidency University, Bengaluru, India 1,3

Department of MCA, Sri Venkateswara College of Engineering & Technology (Autonomous), Chittoor, Andhra Pradesh, India2
Dept of ECE, School of Engineering, and Technology, Amity University Kolkata, India4
Department of Information Technology, VIT University Vellore, India5
Email: [email protected]

Keywords: Cloud-sim tools, amount of work, resource I. INTRODUCTION

Mapping, RFD, PSO, ACO, Provisioning. The critical structures of cloud contain broad network
access, on-demand service, pooling of resources, measured
Abstract services, and rapid elasticity. For example, when we store
Cloud Computing is an innovative model that is changing data or file online, it becomes a service of the cloud. The
the industry procedure of inventing, developing, file may available at any time in the cloud for access and
deploying, order, updating, maintaining, and budget for the user is freed from storing the file on his own storage.
applications and the stock on which they run. It is a Cloud computing is a method that supports easy and broad
technique to increase the resource facility or proficiencies network access where the pool of configuring resources
without financing in a new hardware or licensing new and provide the resources rapidly. There are a lot of
researchers that has proposed different definitions of cloud
software. Due to its attractive features, organizations and
computing. The National Institute of Standards and
individuals are getting involved in cloud technology
Technology (NIST) defines [1] – “The cloud model has
paradigms. However, there are still a few obstacles in four deployment models, five characteristics and three
total deployment of cloud technology commercially, which service models. Cloud computing is a model for enabling
is a challenging task. The complexness of the scheduling convenient, on-demand network access to a shared pool of
job, the resource parameter should be augmented by configurable computing resources (e.g., networks, servers,
decreasing the execution time. Provisioning the resource storage, applications, and services) that can be rapidly
and mapping the workload is a crucial role in improving provisioned and released with minimal management effort
the performance of the complete system. or service provider’ interaction”,
Mapping/Executing the cloudlet should be customer
satisfaction and economical to the provider. Cloud The properties of the deployment model, which consists of
four types such as public cloud, private cloud, hybrid
technology is facing different challenging issues for
cloud, and community cloud. Public Cloud or external
allocating the job in an efficient manner to requested user Cloud applies to familiar people/general people. Public
in the peak time. This thesis is composed of most Cloud is accessible to all at any point in time. It may help
important works like Provisioning and scheduling the people or organization. Private cloud is applicable for
required job, RFD metaheuristics algorithm, Performance business or an organization for security purposes and is
metrics calculation for response time, execution time, and managed by the third party. Hybrid cloud is the grouping
cost. The current research investigation is the job of more than two models (community, private, and public).
Scheduling problem of cloud scheduling paradigm and to And, community cloud is accessible or share by numerous
propose a new approach for provisioning and allocating organization together. Virtualization technique is modest,
the resources using RFD algorithm-based technique. The here various operating systems lane on a single machine
aim of implementing the RFD algorithm in the cloud is to and lots of servers are run at a time. For example, if we are
efficiently allocate the resources and better utilization of using a unique operating system like Windows 10, we can
cloud resources. The proposed effective approach of still operate multiple OS like Vista, XP, and Linux, etc. [2]
resource provisioning and scheduling algorithm is to Nowadays, several IT industry moved to the cloud due to
solve the provision of resource and to execute workload investment profit increases is computing service, it is
efficiently in the cloud paradigm. The theme of research straightforward to access and capital cost-reducing, etc.
contains the proposed River formation dynamics
algorithm which gives a better optimal solution for
allocating the resources in the cloud in Comparing with II. LITRETURE REVIEW
“Particle Swarm Optimization (PSO) and Ant Colony
Optimization (ACO)." The implementation results show A. Provisioning of Resources
decreasing response time, cost, and execution time. The central part is the provisioning of resource detection
and resource selection. According to the availability,
Resource provisioning allows both provider and consumer

Authorized licensed use limited to: AMITY University. Downloaded on June 11,2023 at 11:04:55 UTC from IEEE Xplore. Restrictions apply.
to access the specified resources. Always resource resource [9]. This framework model has improved the
provision happens during the scheduling phase as mapping satisfactory level of customer and provider. Here, the cloud
of resources and workloads; first resource provision will environment scheduling frameworks have developed to
occur then proceeds mapping/scheduling of resources. It route the user request adequate. The scheduling
contains two approaches: architecture is exposed in figure: 1 these set of jobs were
measured as the input. The collection of job request
x Economic approach (cost and time) containing job information, job communicate has arrived
for implementation. The job request contains task id,
x Non-economic approach (FIFO, LB, RR length, and requirements details of the user.
algorithms) [3] [4]

Resource providers will follow some rule before the

provision of resources that is SLA, for implementing the
process should be analysed. Resource provisioning
maintains job allocation reasonably and efficiently. First of
all, the requested user access the resource through the
cloud portal. According to the service level agreement
strategy send execution application to the consumer. The
resource manager (RM) maintains the status of the
resource, SLA, information about the resource, and policy.
Resource manager checks the policy-based availability of
the resource if the resources are available in the repository
then provided to the user. [5]

B. Resource Allocation
Nowadays, to allocating the resources efficiently is most
difficult issues in the computing environment. The details
of resource allocation process is to increase the flexibility
of cloud resource in according to the user constraint. The
most crucial process of allocating the resource is the user
has no requirement for installing any software application
and hardware infrastructure, In Resource Allocation
process has to assign the available resource to a specific
user over the Internet. The resource allocation-based plan Fig. 1. Proposed Scheduling model.
is merely comparing the idea with different parameters the
user request, such as computation resources in the form of According to the portal and web browser submission, it
a lease. The lease is submitted in the form of mode. sends to the resource manager. The resource manager
analyzes all the job and recognizes all the demands
x AR (Advanced Reservation): firstly, resources are (computing, storage, security, and QoS) in details. On
reserved in advance manner, and it allows the accomplishment, the job broker has performed as the
resources to the user in peak time or available middle between the job info and model of resources. The
time. [6] resource modeling is deployed to treat the state and the
x BE (Best Effort): Now Resources are in queue appearances of the resources in the host computer. The job
manager maintains the working process for the job
and peak time of available resources are provision
execution. It sends commands to the scheduling insurance
for the user as soon as possible.
model to schedule all the job based on the collection
x Immediate: This is possible, according to the provided by the job broker hypothesis and the resource
availability of resources. When the user submits a model. The set of a job is delivered to the scheduler. On
request, either it should provide immediately or reaching, the scheduling policy model allocates the
reject. [7] resources to VMs that hold the jobs supported on the RFD
x DS (Deadline Sensitive): DS lease is pre-emptible algorithm and the performance metrics such as execution
and flexible in time constraints. time, expenditure, and response time will be calculated on
x ALT-RA (Alert Time based Resource allocation): the presentation of the scheduling algorithm.
It is based on alert time to assign the available
resources to the user. [8] D. Resource Scheduling Algorithm Framework
The main classification of resource allocation is of two
C. Scheduling Models
types of static resource allocation, and dynamic resource
The primary purpose of the research is to assign the allocation, and the evolution contains RSA (Resource
physical resources to the customer and satisfying the user scheduling algorithm), resource scheduling (RS). For
demand based on SLA. The service request in the cloud optimizing any scheduling algorithm, QoS factors like
paradigm is heterogeneous in nature for mapping the

Authorized licensed use limited to: AMITY University. Downloaded on June 11,2023 at 11:04:55 UTC from IEEE Xplore. Restrictions apply.
 time, budget, and energy show a significant role in values to VM altitude will increases. Cloudlet eroded the
allocating the resources [10][11]. The RSA plays the most adjacent VMs. Finally, the best solution will find in the
fundamental role in providing and scheduling an adequate process by analyzing the path and local cycle will not
resource to a particular user. Because of the dynamic create.
behavior of RSA, Resource is scheduled efficiently. And
the statics resource algorithm is less efficient than dynamic
RSA. Categories of RSA is exposed in above figure :2
Implementation of RSA is designed to avoid both Step1- Number of resource and number of workload
underutilized and over utilized of resources.
Step2:- Initialized Resources [Number of resource list]
Step3:- Initialized workloads [Number of job lists]

D= droplet
N=VM
IN=Initialized VM
DN=Destination VM
At=Altitude of VM
DN=0;

Step4:- While (All the cloudlet did not follow the same
path ()) and (condition are not Match ())
Step5:- Move all the droplet around Vm ()
Fig. 2. Resource scheduling Algorithm Classification. Step6:- Altitude of VM erode the path ()
Step7:- Altitude of all VM deposit sediments ()
Step8:-Analysed the path completely ()
III. PROPOSED SYSTEM
In real life environment to solve optimization problem is Step9:- End While
very difficult because it lies behind the variable, Step10:-Mapping between all workload and resources
constraints, and objectives. So that it is inefficient and
impossible to solve using different analytical methods and
the global optimization problem will be resolved by Fig. 3. Proposed Algorithm.
differential calculus methods. The solution of optimization
problems, some investigator had tried to develop bio-
inspired algorithm that is based on heuristics or
Metaheuristics formulation to judge the best or linear
optimal result for the problems. The global optimum can B. Experimental Evaluation
be exclusive if the formula is fit to confront the enormous
The RFD algorithm is achieved quick results for any
place.
workload instances, and table:1 is showed specific
A. The Algorithm parameters of RFD. The projected algorithm is operated in
Here VM(k): is the set of VM. VM(i) is connecting VM, 500 iterations for executing all resources efficiently.
and then all the workload will move until all executed.
Here according to an iteration of workload execution, The experiments were distributed in the simulation of
calculated the value of time, response time, and cost. And cloudsim toolkits provisioning and implementing
the negative gradient value of decreasing gradients (i.j) resources. A single machine does the development of the
between two VMS will calculate in following equation 4. test with the same parameter for all the algorithms.

When the new slope creates between initial to destination

VMS, altitude will increases. Adding some sediments

Authorized licensed use limited to: AMITY University. Downloaded on June 11,2023 at 11:04:55 UTC from IEEE Xplore. Restrictions apply.
 IV. RESULTS
Test Case 1: Execution Time with the number of
resources/number of workloads: The simulation results
displayed in figures: 4 that for small workloads the
execution time of RFD is 10.48% lesser than PSO and 27
.54% lesser than ACO 11.26% lesser than PSO and 49.75%
lesser than ACO. . Like that, for an extra high workload, the
execution time of RFD is 15.33% lower than PSO, and
78.92% smaller than ACO. When execution is more with
the number of workloads also increases.

Table1: Parametric resource table.

C. CloudSim Simulator
There are different characteristics to design the cloudsim
toolkits such as cloudlet, cloudlet scheduler, host, VM,
network topology. Cloudlet is providing a different model Table3: Average Execution time (50 drops).
for generating workload, and numerous task is executing
competently. Cloudlet scheduler contains different policy
to implement cloudlet proficiently. The host is nothing but
a physical machine [12]. It provides CPU, memory,
bandwidth, etc. To running the user application and policy,
VM is required. Network topology contains different
parameters and monitors the delay of the resources. The
simulated cloud environment mainly includes a user,
datacenter, and broker. The delay of resource happened in
runtime allocating resource and initialization of all
resource at a time. It is a customized tool, where the
developer can provision and allocate resource with a
different policy without spending any cost and also check
the performance of deployment [13]. The feature of
cloudsim provides a different platform of scheduling,
allocation policies datacenter modeling, and service
broker. It offers enormous infrastructure to implement the Fig. 4. Execution Time Comparison with the change of resources
(50 drops).
resources in a single node and virtualization environment
to manage all resources. Table: 2 shows the different host Test Case 2: Execution Cost with the number of
and VM parameter to implement our proposed algorithm. resources/number of workloads: The simulation results
[14] displayed in figures: 5 that for small workloads the
execution Cost of RFD is 13.51% lesser than PSO and
4.96% lesser than ACO. Like that, for an extra high
workload, the execution cost of RFD is 10.44% lesser than
PSO, and 62.95% lesser than ACO. When execution cost is
more with the number of workloads also increases. In
figure (5.12, 5.13 and 5.14) execution time is comparing
with resources (15-150) with different size of drops (50,
100, and 150), 50 drops the execution cost of RFD is
24.93% lesser than PSO and 56.29% lesser than ACO.

Table 2: VM and Host resources.

Authorized licensed use limited to: AMITY University. Downloaded on June 11,2023 at 11:04:55 UTC from IEEE Xplore. Restrictions apply.
 Table 4: Average Execution Cost (50 drops).

Fig. 6. Response Time Comparison with the change of resources

(50 drops).

V. CONCLUSIONS AND FUTURE WORKS

The proposed algorithm, in comparison, uses metaheuristics
algorithm (ACO and PSO) allocating the resources
efficiently. In conclusion, the proposed methods contain the
RFD algorithm for the best optimal solution for effective
provision and proficiently scheduling for resources with
decreased execution time, cost, and minimized response
time. However, the Proposed RFD task scheduling
algorithm offers a better optimal solution for the
provisioning and allocating of numerous resources
Fig. 5. Execution Cost Comparison with the change of resources
(50 drops).
effectively. In the future, we will implement different
parameters of QoS (Quality of services) metrics, such as
testing time, reliability, serviceability, throughput, and
transit delay. The proposed algorithm can be further
developed automatic management for cloud resources using
SLA agreements. And in future, the proposed RFD
algorithm will be extended to (LB-RFD) Load balancing
RFD for allocating more number of resources. The
proposed system has implemented in cloudsim simulation
tools, in future same provisioning parameter will implement
in the real-time cloud environment.

REFERENCES

[1] Mell, Peter, and Tim Grance. "The NIST definition of cloud
Table 5: Parametric resource table.
computing," 2011.
[2] Nayak, Nihar Ranjan, and S. Bhubaneswari. "RFD based on
Test Case 3: Response Time with the number of Resource Scheduling Technique in cloud Computing,"2019.
resources/number of workloads: The simulation results [3] Dasgupta, Gargi B., et al. "Qos-graf: A framework for qos based grid
displayed in figures (5.15, 5.16, 5.17 and 5.18) that for resource allocation with failure provisioning." 200614th IEEE
International Workshop on Quality of Service. IEEE, 2006
small workloads the Response time of RFD is 20.92% [4] Gajbhiye, Amit, and Shailendra Singh. "Resource Provisioning and
lesser than PSO and 28.21% lesser than ACO. Like that, for Scheduling Algorithm for Meeting Cost and Deadline-Constraints of
an extra high workload, the Response time of RFD is Scientific Workflows in IaaS Clouds." arXiv preprint arXiv:1806.02397 ,
18.51% lesser than PSO, and 23.94% lesser than ACO. 2018.
[5] Aron, Rajni, and Inderveer Chana. "Formal QoS policy based grid
When response time is more with the number of workloads resource provisioning framework." Journal of Grid Computing 10.2 249-
also increases. In figure (5.19, 5.20 and 5.21) Response 264, 2012.
time is comparing with resources (15-150) with different [6] He, Eric, Xi Wang, and Jason Leigh. "A flexible advance reservation
size of drops (50, 100, and 150), 50 drops the response time model for multi-domain WDM optical networks." 2006 3rd International
Conference on Broadband Communications, Networks and Systems.
of RFD is 13.26% lesser than PSO and 37.90% lesser than IEEE, 2006.
ACO. [7] Toosi, Adel Nadjaran, et al. "Resource provisioning policies to
increase iaas provider's profit in a federated cloud environment." 2011

Authorized licensed use limited to: AMITY University. Downloaded on June 11,2023 at 11:04:55 UTC from IEEE Xplore. Restrictions apply.
IEEE International Conference on High Performance Computing and [12] Goyal, Tarun, Ajit Singh, and Aakanksha Agrawal. "Cloudsim:
Communications. IEEE, 2011. simulator for cloud computing infrastructure and modeling." Procedia
[8] Nayak, Nihar Ranjan, and S. Bhubaneswari. "An RFD Algorithm Engineering 38 ,3566-3572, 2012.
Based on Resource Provisioning and Scheduling Proficiency in Cloud [13] Rosenberger, Leland, et al. "XPS analysis of aluminum nitride films
Paradigm,"2018. deposited by plasma source molecular beam epitaxy." Surface and
[9] Yoo, Seunghwan, and Sungchun Kim. "SLA-aware adaptive Interface Analysis: An International Journal devoted to the development
provisioning method for hybrid workload application on cloud and application of techniques for the analysis of surfaces, interfaces and
computing platform." Proceedings of the international multiconference thin films 40.9 ,1254-1261, 2008.
of engineers and computer scientists. Vol. 1. 2013. [14] Buyya, Rajkumar, Suraj Pandey, and Christian Vecchiola.
[10] Nepal, Surya, and Mukaddim Pathan, eds. Security, privacy and "Cloudbus toolkit for market-oriented cloud computing." IEEE
trust in cloud systems. Springer Berlin Heidelberg, 2014. International Conference on Cloud Computing. Springer, Berlin,
[11] Zare, Javad, et al. "Resource scheduling in mobile cloud computing: Heidelberg, 2009.
taxonomy and open challenges." 2015 IEEE International Conference on
Data Science and Data Intensive Systems. IEEE, 2015.

Authorized licensed use limited to: AMITY University. Downloaded on June 11,2023 at 11:04:55 UTC from IEEE Xplore. Restrictions apply.