Journal of Ambient Intelligence and Humanized Computing (2023) 14:2185–2197

https://doi.org/10.1007/s12652-021-03429-w

ORIGINAL RESEARCH

ABSO: an energy‑efficient multi‑objective VM consolidation using

adaptive beetle swarm optimization on cloud environment
B. Hariharan1 · R. Siva1 · S. Kaliraj2 · P. N. Senthil Prakash3

Received: 11 December 2020 / Accepted: 5 August 2021 / Published online: 15 August 2021
© The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2021

Abstract
Cloud computing is a powerful way to provide a suitable platform for data centers and to store data. Along with the so many
benefits, there are still some management issues that need to be investigated. Although cloud computing seems to be a very
attractive implementation it is facing incredible energy consumption and costs concerns. To avoid energy consumption, a
VM consolidation and migration approach is introduced. The main objective of VM consolidation is to perform more jobs
while consuming less amount of power. To achieve this, in this paper multi-objective energy-efficient VM consolidation
using adaptive beetle swarm optimization (ABSO) algorithm is proposed. The proposed ABSO is a hybridization of particle
swarm optimization (PSO) and Beetle swarm optimization (BSO).The proposed method presented with efficient solution
representation, derivation of efficient fitness function (or multi-objective function) along with PSO and BSO operator. The
effectiveness of the approach is analyzed based on the different evaluation measures and effectiveness is compared with
different methods. From the results, our proposed approach consumes only 8.234 J energy for scheduling 100 tasks which
are 10.616 J for BSO-based VM consolidation, 11.754 J for PSO-based VM consolidation, and 13.545 J for GA-based VM
consolidation.

Keywords VM migration · Consolidation · Load balancer · Multi-objective function · Particle swarm optimization · Beetle
swarm optimization

1 Introduction deployment models. Cloud computing helps users to save

infrastructure and maintenance costs (Thakur et al. 2013).
Cloud computing has been popular in the IT industry since Users receive scalability, reliability, and mobile accessibility
2008. It offers computing services as a "pay-as-you-go" facilities from cloud computing services. Various companies
model, which enables utility computing, such as gas, elec- focus on their innovation creating business values by over-
tricity, water, and telephone utilities. It could be referred to looking the low-level hardware and software infrastructure
as the 5th utility. Softwares a service (SaaS), Infrastructure setup. The process of moving all computing services into the
as a Service (IaaS), and Platform as a Service (PaaS) are cloud is slowed down by various open challenges. Security,
the commonly known cloud computing services (Suakanto privacy, energy efficiency, and common standards restrict
et al. 2012). These services are deployed by following vari- many companies from migrating their computing services
ous deployment models. Depending on the requirements into the cloud (Ahmad et al. 2015).
public, private, commodity and hybrid are four major cloud Virtualization is the one of center innovation in dis-
tributed computing. Typically, data centers utilize the
* B. Hariharan deployed servers to offer various types of assistance and it
[email protected] will underutilize the server and maximize the energy costs.
An extraordinary number of VMs are arbitrarily conveyed
1
Department of School of Computing, SRM Institute in data centers to offer the assistance for end- users as indi-
of Science and Technology, Kattankulathur, Chennai, India
cated by the necessary assets (Elgendy et al. 2020). Due
2
Department of Software Engineering, SRM Institute to the rapid development of cloud computing technology
of Science and Technology, Kattankulathur, Chennai, India
companies are using a lot of data centers. As a result, the
3
Department of Computer Science and Engineering, RMK problem of high power consumption is becoming more and
College of Engineering and Technology, Chennai, India

13
Vol.:(0123456789)
2186 B. Hariharan et al.

more acute (Madhan and Srinivasan 2014; Hariharan and • The paper proposes adaptive beetle swarm optimization
Raj 2020). Energy consumption is an important issue for (ABSO) algorithm by integrating BSO with PSO for
data center operators because it increases cost and envi- selecting migrated PM. The algorithm can effectively
ronmental issues. Therefore, reducing data center energy find proper an underutilized PM and migrate out all of
consumption is an economic incentive for data center the VMs running on the PM, then switches the PM to
operators. And it will lead to a more stable environment sleep mode for power saving.
as it helps to reduce the global ­CO2 pathway. Amazon esti- • For experimentation, different kinds of PMs are consid-
mates that data center data costs about energy consump- ered and they can be applied to large heterogeneous cloud
tion, including direct consumption (approximately 19%) data centers.
and cooling infrastructure (approximately 23%) (Sharifi • The performance of the proposed approach is analyzed
et al. 2012). Energy consumption is constantly increasing in terms of different metrics and the effectiveness of the
in big data centers like Google or Facebook. E.g. Total proposed method performance is compared with other
energy consumption is increasing due to the expansion of algorithms.
the ability to replace older hardware with higher energy
efficiency or the introduction of efficient cooling systems The rest of the paper is organized as follows: Sect. 2 pro-
to meet the increasing user demand (Khani et al. 2015). vides a literature review of the proposed VM migration and
VM migration has pulled in impressive consideration consolidation. Section 3 describes the problem definition
from data center directors as of late. Industry items such as and solution framework and Sect. 4 shows the proposed sys-
XenMotion and VMotion have been actualized as work in tem model. The proposed methodology is given in Sect. 5
apparatuses in their virtualization stages (Beloglazov and and the analysis of the experimental results is given in
Buyya 2012; Elgendy et al. 2021). VM migration includes Sect. 6. Finally, conclusions are made in Sect. 7.
three central questions of when and where a VM should be
relocated (Hariharan and Raj 2019; Yadav et al. 2018). The
essential for migration is to recognize which VM should 2 Literature review
be moved away from an overloaded PM. It might bring
about various levels of effect on the framework execution Numerous researchers had been developed virtual machine
of the overloaded PM. Notwithstanding, during the migra- consolidation to reduce energy consumption. Among them,
tion and consolidation, a few issues should be thought of: some of the works are analyzed here; Elshabka et al. (2020)
(1) when a host is considered as being delicately stacked had developed security-aware dynamic VM integration.
or daintily stacked? When a host should be optimally Their approach consists of two modules, the Security Moni-
stacked or intensely stacked? (2) Which VMs should be toring Module (SMM) and the Security-Awareness Dynamic
relocated? (3) Where should VMs be put?. Nowadays, a VM Integration (SDVMC) module. Restructuring results
lot of methods and algorithms had developed for energy- demonstrated that by using their strategy with RITH 0.8
efficient VM consolidation. Stochastic algorithms are the effects on safety improvement, the general risk was reduced
usual approaches for solving multi-objective formulations from 2 to 5% without a negative effect on energy use or
of the VM consolidation problem (Gao et al. 2013; Zheng QoS. The RITH value was below 0.1 means, the method
et al. 2016). We extend the VM consolidation problem produces a maximum energy consumption with insignificant
with new objectives and propose a multi-objective func- security improvement. Due to this reason, the method should
tion that outperforms the existing approaches. be further improved. Witanto et al. (2018) had developed
The main objective of the proposed VM consolidation is a neural network-based adaptive selection of VM consoli-
to migrate and place the VMs in a suitable PM and switch- dation. Here, they choose a proper algorithm based on the
ing unused PMs to save energy. To improve the energy cloud provider’s goal priority and environment parameters.
in the data center, the consolidation approach is utilized. By choosing the optimal VM, the NP-hard problem was
Virtual machine migration is a key stage of consolidation. adapted. Therefore, for selecting optimal VM, optimization
In this paper, we propose an ABSO algorithm for VM algorithms are suggested.
migration and consolidation. To achieve energy-efficient Moreover, Mohiuddin and Almogren (2019) had aimed
consolidation multi-objective function is designed. The to develop VM consolidation based on workload. This tech-
main contribution of the proposed methodology is given nique was utilized to switch the inert actual workers into
below; hibernation mode, bringing about decreased force utiliza-
tion. Compared to another method, this method saved 33%
• We design a multi-objective function (MOF) based on of energy during the process. Similarly, Singh and Singh
VM migration cost, memory utilization, and energy (2021) had developed geometric programming-based multi-
parameters for selecting a suitable PM for consolidation VM migration, which assigns compression rates to each

13
ABSO: an energy‑efficient multi‑objective VM consolidation using adaptive beetle swarm… 2187

VM to minimize total migration time. This paper clearly of number of shifts, number of sleeping BMs, number of
showed the memory summary, with copy front displace- SLA violations and power consumption. This system needs
ment, enhancing the live migration exhibition of numerous to be improved.
VM by minimizing the migration time and idle time. Also,
the compression ratio is assigned variable according to the 2.1 Motivation
available network bandwidth. This method was an effec-
tive one but migrating the VMs across WAN is a difficult By providing a suitable platform for data centers, cloud com-
one. Similarly, Ding et al. (2020) had aimed to develop VM puting has become one of the most effective methods for
consolidation model by improving power. To approve the data storage. Despite the many merits of cloud computing,
unwavering quality and adaptability of the proposed tech- some management issues such as energy consumption and
nique, they played out a trial assessment in both genuine and cost still need to be explored. To serve customers, cloud data
mimicked conditions. This system produces the better results centers require thousands of server machines to operate. It
but the robustness of this system was very low. consumes 80% of the total power, even at 20% of server
Saeedi and Shirvani (2021) had aimed to develop VM usage. Not only will it increase operating costs, but higher
consolidation based on optimization approach. For achieve energy consumption will greatly emit carbon. To avoid this,
this objective, simulated annealing (SA) is developed. This cloud providers are trying to find a better way to restructure
algorithm tries to provide the additional server with mini- the IT infrastructure to reduce power outages. One of the
mum energy consumption. The efficiency of this approach methods for integrating VM into a physical server to reduce
was analysed based on four algorithms. Zhang et al. (2020) power consumption is called virtualization integration. Now-
had aimed to introduce load balancing on cloud with IoT. adays, a lot of researchers had a focus on energy-efficient
Here, they clearly explained the security and load balanc- VM consolidation. In Saxena et al. (2021), the author had
ing process. In, Ghetas (2021), had aimed to develop Mon- developed an ant colony optimization algorithm based on
arch Butterfly Algorithm based VM placement. To achieve VM consolidation. This method was failed to maintain the
this concept, they introduced a multi-objective fitness this energy during the VM migration. Moreover, the ant colony
method was maximizing packaging efficiency and reduce the optimization algorithm was mostly dependant on sequences
number of active physical servers. Sayadnavard et al. (2021) of random decisions and uncertain time to convergence.
had aimed to develop a reliable dynamic VM consolidation. Due to this reason, this method attained a poor outcome. To
Here also they designed a multi-objective fitness function. overcome the problem, in this paper hybrid approach-based
This method was save large amount of energy compared to energy-efficient VM consolidation is proposed.
the other approaches. Similarly, multi-objective based load
balancing on cloud had explained in Saxena et al. (2021).
this method was proposed to minimize the network traffic 3 Design of multi‑objective fitness function
and power consumption with datacenter. (MOF)
Moreover, Paulraj et al. (2018) had introduced a resource-
aware VM migration procedure. The assessment results During the consolidation process, tasks are optimally
show that the proposed strategy outflanks the condition of assigned to the VM to reduce power consumption. In this
workmanship procedures regarding the number of migra- paper, a MOF is designed to detect the optimal VM. MOF
tions, energy usage, and migration time. This method should is a hybridization of three parameters: VM migration cost
be improved. Huang et al. (2011) have provided an elec- ­(CM), memory usage ­(UM), and energy ­(EC).
tricity consumption estimate of the effects of direct migra- Let us assume the cloud computing network which}con-
tion of VMs. The results show that the power overlap of sists of M number of PMs PM = PM1 , PM2 , … , PMN and
{

the migration was much lower in the context of the use of each PM consist of N number of VMs.{The aim of this } paper
the integration strategy than in conventional deployment is to optimally assign the taskTi, T = T1 , T2 , … Ti to VM
without the use of integration. Their results are based on a to reduce the power consumption and cost. Each VM has
conventional physics server whose power was a linear model a different quantity of CPUs, varies sized of memory and
of CPU usage percentage. Moreover, Aryania et al. (2018) bandwidth. Likewise, each task has a different size, energy,
had developed energy-awareness VM integration using the and memory. To save energy during the task scheduling
ant colony method. This significantly reduces the number process, we should switch off (sleep mode) the maximum
of displacements and the number of active physics engines, number of PMs. The PM is overloaded means, it takes a
resulting in a reduction in the total energy consumption of lot of time to complete the task. If the PM is underloaded
the data centers. Simulation of random workload in differ- means, we can lose a lot of energy. In this, each PM satisfies
ent situations demonstrates that the proposed algorithm sur- the load condition is an important problem of consolidation.
passes the sophisticated VM integration algorithm in terms The designed MOF is given in Eq. (1).

13
2188 B. Hariharan et al.

Fitness Function = Min 𝛼1 EC + 𝛼2 CM + 𝛼3 U M (1) VM consolidation. To solve the problem, in this paper
[ ( ) ( ) ( )]
hybridization approach-based VM consolidation is
Energy consumption (EC) The first parameter of the proposed.
objective function is EC. The minimum value of EC is
should be a good system. The EC of the system is calcu-
lated simultaneously by the total Euclidean distance (ED) 4 System model of VM consolidation
of all active PMs. Here, small ED is considered the best
performer. When no work is done on the relevant PM, Cloud computing is a growing paradigm that utilizes com-
the PM is switched
( ) off. For every active PM, the energy- puter resources in online distribution services. Cloud data
efficient factor E𝛿 is evaluated which is given in Eq. (2). centers are designed to make efficient use of cloud resources
√ such as networks, storage, services, applications, and serv-
ers. One of the most important issues for the operation and
√ d ( )2
√∑
E𝛿 = √ Uj − uUbestj (2) maintenance of cloud data centers is power consumption.
Cloud applications use high amounts of energy, high oper-
j=1

where, j → Resources namely CPU and memory, ­Uj → Cur- ating costs, and carbon emissions for the environment. To
­ bestj → Best energy efficiency in
rent resource utilization, U serve customers, thousands of server machines operate
each PM. in the cloud computing environment. Even at 20% usage,
The system energy efficiency ET at time T is repre-
( ) servers use 80% power. In addition to increasing operating
sented as the total ED of all active PM at this time. The costs, higher EC also causes greater emissions of carbon.
ET is given in Eq. (3). To avoid this, cloud providers are trying to find an effective
mechanism by which IT infrastructure can be restructured to
reduce energy depletion. One of the methods for integrating
∑
ET = E𝛿 (3)
the number of VM in a physics server to reduce EC is called
The total energy efficiency of the system within T is virtualization. The integration process is given in Fig. 1.
given by (4) VM integration aims to replace VMs with suitable PMs
N
and replace unused PMs to conserve energy. Optimal VM
E =
∑
ET (4) placement achieves energy efficiency in the cloud and cre-
T=0 ates data centers in the cloud model to stay energy alert. In
this VM integration, the client sends the request to the cloud
Migration cost MC is the second parameter of VM con- network as tasks, and each task is transmitted approximately
solidation. The migration cost of VM is expanded when to the VM. The proposed VM consolidation system model is
the quantity of migrations increases. The decent consolida- given in Fig. 2. The cloud network consists of different com-
tion framework ought to keep up the least migration. The ponents namely, physical machines (PMs), virtual machines
evaluation of the MC of the whole Cloud arrangement is (VMs), and load balancers (LB). The PM controls the VM
determined utilizing Eq. (5). and loads balance is used to check the availability of PM.
N ( ) Energy consumption is a major role in the proposed VM
1∑ Number of migration in VMs
Migration Cost (MC) = consolidation. The resources namely, CPU, memory, band-
N i=1 Total number of VMs
width, and energy play a role in energy consumption in the
(5)
cloud data center.
Memory utilization Memory usage is the third parame-
ter of MOF. The calculation of the MU of the whole Cloud
arrangement is determined utilizing Eq. (6). 5 Proposed energy‑efficient VM
memory(M) consolidation using ABSO
1
= The main objective of the proposed methodology is to
PM × VM
[P V develop an energy-efficient virtual machine migration and
)]
1 CPU utilizedij memory utilizedij consolidation based on ABSO. The proposed ABSO algo-
M M
(
∑ ∑
+
2 CPUij memoryij rithm is the hybridization of PSO and BSO. For scheduling,
i=1 j=1
(6) the proposed ABSO migrates the VM in the under-loaded
PM to the overloaded PM and switch off the unwanted PM to
In Eq. (1), the objective function of our research is men-
save energy. The overall structure of the proposed methodol-
tioned. From the formula, we can see that the problem is
ogy is given in Fig. 2.

13
ABSO: an energy‑efficient multi‑objective VM consolidation using adaptive beetle swarm… 2189

Fig. 1  Consolidation process

Step 1. Solution encoding: Initialization solution encoding Fitness Function = Min 𝛼1 EC + 𝛼2 C M + 𝛼3 U M

[ ( ) ( ) ( )]
is the first step for optimization problems. Initially, the (7)
task is randomly assigned to the PM. The solution con-
Step 3. Updation using ABSO algorithm After the fitness
sists of the number of tasks, VM, and PM. The task has
calculation, the solutions are updated using the ABSO
three different characteristics namely, size, energy, and
algorithm. ABSO is a combination of BSO and PSO.
memory. Similarly, each PM has several VM, memory,
BSO is an intelligent algorithm that mimics the beetle
and bandwidth. Based on the characteristics, the tasks
hunting process. In BSO, there is a beetle that senses
are assigned to the VM while minimizing cost, time,
the intensity of the odor emitted by the left and right
and number migrations. For example, we consider, the
antennae from the food. The beetle learns the differ-
solution consists of three PM and each PM consists of
ent senses of the two antennas to determine flight and
three VM. The solution encoding format is given in
finally finds food. The BSO algorithm has been proven
Fig. 3.
to have good optimal speed and accuracy when used
for low-dimensional optimization problems. How-
In Fig. 3, the seven tasks are assigned to VM11 , VM31 ,
ever, when solving high-dimensional problems, the
VM12 , VM12 , VM22 , VM13 , and VM23 respectively.
algorithm can easily get stuck locally optimized. To
enhance the optimization efficiency of BSO, we call
Step 2. Fitness calculation After the solution encoding
the Particle Mass Optimization (PSO) algorithm with
process, the fitness of each solution is calculated. The
BSO as ABSO. At ABSO, first, standard PSO is used
fitness function design is already explained in Sect. 4.

13
2190 B. Hariharan et al.

Fig. 2  Overall structure of proposed methodology

PM1 PM2 PM3

VM 11 VM 12 VM 31 VM 12 VM 22 VM 32 VM 13 VM 23 VM 32

1 0 1 1 1 1 1 1 0

Fig. 3  Solution encoding

for particle position and velocity updation. Particles a beetle independently applying a local search with a
learn oncology information and group optimization for BSO algorithm. In the updation process, initially, the
evolution. Then each particle in the swarm is taken as

13
ABSO: an energy‑efficient multi‑objective VM consolidation using adaptive beetle swarm… 2191

PSO algorithm is applied. The PSO-based velocity and Xrst+1 = Xrst + Qtjs ∗ d∕2
position updations are given in Eqs. (8) and (9). (14)
Xrst+1 = Xrst − Qtjs ∗ d∕2
( )
Vjt+1 =w ∗ vtj + 𝜑1 × rand1 × Pbestt − Xjt
j Using this procedure, the optimal solutions are selected.
(8)
The selected optimal solution is given to the consolidation
( )
× Pbestt + 𝜑2 × rand2 Gbestt − Xjt
j j
process.

Xjt+1 = Xjt + Vjt+1 (9) Step 4. Step 4. VM consolidation Once the optimal solution
is attained using ABSO, based on the solutions, the
where, j and w denotes the particle at jth position and tasks are assigned to each VM. After the scheduling,
inertia weight. Then 𝜙 1 and 𝜙2 are the learning rates. the load of each PM is checked. The threshold value
The rand1 and rand2 are referred to as the random num- T H is used to calculate the load of the PM. If the load
bers that are uniformly distributed in the range [0, 1]. value ­(LV) increases the T H means, the correspond-
The Vjt is denoted the maximum and minimum range ing PM is overloaded and we should migrate the PM.
of [vmax , vmin ] velocity. Similarly, if the PM is underloaded with minimum load
means, we migrate the VM into another PM and switch
After the PSO-based updation, the solutions are updated off the PM.
using BSO. In this, the particles in the swarms are con- Step 5. Stopping criteria The algorithm terminates its per-
sidered independent beetles. The BAS algorithm is used formance when the best fitness value is selected. Once
to ensure that beetle individuals independently learned the best fitness is attained by means the solution is
information about their surroundings for local search. given to the consolidation process. The flowchart of
With this learning mode, people are moving towards an ABSO based consolidation is given in Fig. 4.
optimal solution to the problem. The mathematical model
for simulating its behavior is as follows;

Xjst+1 = Xjst + 𝜆Qtjs + (1 − 𝜆)𝜀tjs (10) 6 Result and discussion

where; t → Current number of iterations, Qjs → Speed of The results obtained by the proposed ABSO based virtual
beetles, 𝜀js → Beetle position movement. machine migration and consolidations are presented here.
The speed of the beetle is calculated as follows; The proposed approach is implemented using Java (jdk1.6)
( ) ( ) with cloudSim devices and a progression of experiments was
Qt+1
js
= 𝜔Q t
js
+ c1 r1 U t
js
− Pt
js
+ c2 r2 U t
gs
− Pt
gs (11) performed on a PC with Windows 7 Operating framework
at 2 GHz dual-core computer with 4 GB RAM running a
where, c1 , c2 → Two optimistic constants, r1 , r2 → The arbi- 64-bit version of Windows 2007. The proposed methodology
trary functions inside the range [0,1], 𝜔 → Inertia weight. is analysed in terms of migration cost, memory utilization,
This paper adopts the strategy of decreasing iner- and energy consumption.
tia weight, and which is mathematically represented as
follows;
6.1 Experimental results
𝜔 − 𝜔max − 𝜔min
𝜔 = max ∗t (12)
T The purpose of the proposed method is to assign the task to
the corresponding VM to minimize energy consumption. To
where, 𝜔min , 𝜔max → denotes the minimum as well as the
prove the effectiveness of the proposed approach, we com-
maximum value of 𝜔 , T,t → denotes the present number
pare our algorithm with different methods. For experimental
of iterations as well as the maximum number of iterations,
analysis, two configurations are used such as (i) PM = 10,
Here, the 𝜉 function defines the incremental function
VM = 20, and 50 tasks, and (ii) PM = 25, VM = 60, and 100
which is mathematically represented as follows;
tasks.
𝜉jst+1 = 𝛿 t ∗ Qtjs ∗ sign(f (Ptrs ) − f (Ptrs )) (13)
(i) (i) Experimental results based on first configuration
In this step, we extend the update solution to a high (PM = 10, VM = 20 and 50 task)
dimension. here 𝛿 indicates the step size. The updation
function of right and left antenna calculation is given The main objective of this section is to assign the fifty
below; tasks to corresponding virtual machines and switch off the

13
2192 B. Hariharan et al.

Fig. 4  Flow chart of proposed

unwanted PMs to save energy. Figures 5, 6 and 7 shows 12,326 MB, which is 172,36 MB for BSO-based VM con-
the performance of the proposed approach using this solidation, 18,526 MB for PSO-based VM consolidation,
configuration. and 20,524 MB for GA based VM consolidation. This case
In Fig. 5, the efficiency of the recommended model is the same for all the remaining iterations. Therefore, it
is discussed in terms of memory utilization. Here, the can be concluded that the recommended model has out-
x-axis represents the number of iteration and the y-axis performed the GA, PSO and BSO approaches. This is due
represents the memory usage. To prove the effectiveness to, hybrid algorithm (ABSO) utilized for load balancing in
of the recommended model, we compare our proposed VM consolidation. This algorithm overcomes the difficul-
ABSO algorithm with BSO-based VM consolidation, ties present in the individual PSO and BSO. In Fig. 6, the
PSO-based VM consolidation, and GA-based VM con- efficiency of the recommended model is analyzed in terms
solidation. When analyzing Fig. 5, in iteration 30, the of energy consumption by a varying number of iterations.
proposed approach reaches a minimum memory usage of When the energy consumption is low, then that method can

13
ABSO: an energy‑efficient multi‑objective VM consolidation using adaptive beetle swarm… 2193

Fig. 5  Comparative analysis

based on memory utilization 25000

20000

Memory usage (MB)

15000

10000

5000

0
10 20 30 40 50
GA 21768 22923 20524 21923 23567
PSO 19467 20634 18526 19452 20175
BSO 17682 18423 17236 18975 19432
ABSO 13456 14432 12326 14643 15428

Fig. 6  Comparative analysis

based on energy consumption 12
Energy consumption (J)

10

0
10 20 30 40 50
GA 10.643 10.854 10.545 10.934 11.234
PSO 8.324 8.643 8.254 8.845 8.934
BSO 7.345 7.532 7.216 7.734 8.123
ABSO 6.325 6.634 6.235 6.745 6.953

Fig. 7  Comparative analysis

based on energy consumption 6

4
Cost ($)

0
10 20 30 40 50
GA 4.4 4.6 4.2 4.7 4.9
PSO 3.2 3.4 3.1 3.6 3.8
BSO 2.8 2.9 2.7 2.9 3.2
ABSO 1.5 1.6 1.4 1.8 1.9

13
2194 B. Hariharan et al.

be concluded as the efficient one. When analyzing Fig. 6, (ii) Experimental results based on second configuration
our recommended model is taken the minimum energy of (PM = 25, VM = 60 and 100 task)
6.235 J for scheduling 50 tasks on 20 VMs. The methods,
In this section, the experimental results are carried out
PSO and BSO are more are less perform equally. Simi-
based on the second configuration. For the second configu-
larly, in Fig. 7, the efficiency of the recommended model
ration, 100 tasks are assigned to 60 VM, and extra PMs are
is analyzed in terms of cost. If the number of migration
switched off to reduce the energy consumption. The results
increases means cost also increases. The minimum cost
obtained by this configuration are given to Figs. 8, 9 and 10.
should be a good system. When analyzing Fig. 7, recom-
In Fig. 8, the efficiency of the recommended model is
mended model need 1.4$ for scheduling 50 tasks on 20
discussed using memory utilization. Here, the x-axis rep-
virtual machine. We can’t able to assign the entire task
resents the iteration and the y-axis represents the memory
on PM. The PM is overloaded means all the tsk assigned
usage. When analyzing Fig. 8, the recommended model
particular PM is collapsed and energy is also loosed. From
taken 14326 MB memory for scheduling 100 tasks which are
the results, it’s clear that the proposed approach attained
19236 MB memory for BSO-based scheduling, 21526 MB
better results compared to another algorithm. This is due
for PSO-based scheduling, and 23524 MB for GA-based
to ABSO based load balancing.

Fig. 8  Comparative analysis

based on memory utilization 35000
30000
Memory usage (MB)

25000
20000
15000
10000
5000
0
10 20 30 40 50
GA 24768 26923 23524 28923 30567
PSO 22467 24634 21526 26452 28175
BSO 20682 22423 19236 23975 25432
ABSO 15456 16432 14326 18643 19428

Fig. 9  Comparative analysis

based on energy consumption 16
14
Energy consumption (%)

12
10
8
6
4
2
0
10 20 30 40 50
GA 13.643 13.854 13.545 13.934 14.234
PSO 11.824 11.943 11.754 11.945 12.234
BSO 10.745 10.932 10.616 10.934 11.323
ABSO 8.321 8.633 8.234 8.746 8.957

13
ABSO: an energy‑efficient multi‑objective VM consolidation using adaptive beetle swarm… 2195

Fig. 10  Comparative analysis

based on migration cost 7
6
5
4

cost ($)
3
2
1
0
10 20 30 40 50
GA 5.6 5.9 5.5 5.7 6.1
PSO 4.2 4.4 4.1 4.6 4.8
BSO 3.8 3.9 3.7 3.9 4.2
ABSO 2.5 2.6 2.4 2.8 2.9

scheduling. From the figure, it’s clear that the recommended Fig. 10, the proposed method taken less cost for scheduling
model attained better results compared to GA, PSO, and 100 tasks compared to other algorithms. From the results,
BSO-based scheduling. In Fig. 9, the performance of the it has been clear that the proposed ABSO algorithm has
recommended model is analyzed in terms of energy con- performed much better than GA, PSO, and BSO algorithms
sumption. Energy is an important parameter for consolida- based on energy consumption, memory utilization, and
tion and migration. The minimum energy consumption is migration cost for various configurations.
a good system. When analysing Fig. 9, our recommended
(iii) Time complexity analysis
model is consumes 8.234 J energy for scheduling 100 tasks
which are 10.616 J for BSO-based scheduling, 11.754 J for One of the main parameters of VM consolidation is time.
PSO-based scheduling, and 13.545 J for GA-based schedul- The total time taken for completing a task is called compu-
ing. From the results, it’s clear that the proposed approach tation time. We analyze the performance of the proposed
is better than other algorithms. The comparative analysis approach using the time is given in Fig. 11.
based on migration cost is given in Fig. 10. When analyzing

Fig. 11  Computation time

analysis 800

700
Computation time (ms)

600

500

400
50 Task
300 100 Task
200

100

0
GA PSO BSO ABSO
Methods

13
2196 B. Hariharan et al.

When analysing Fig. 11, the proposed approach taken a Ding W, Luo F, Han L, Gu C, Lu H, Fuentes J (2020) Adaptive virtual
minimum time of 340 ms for completing fifty tasks which machine consolidation framework based on performance-to-power
ratio in cloud data centers. Futur Gener Comput Syst 111:254–270
is 310 ms for GA-based consolidation, 320 ms for PSO- Elgendy IA, Zhang WZ, Zeng Y, He H, Tian YC, Yang Y (2020) Effi-
based consolidation, 300 ms for BSO-based consolidation. cient and secure multi-user multi-task computation offloading for
Compared to GA, PSO, and BSO, our proposed method mobile-edge computing in mobile IoT networks. IEEE Trans Netw
is taken more time for consolidation. This is due to our Serv Manag 17(4):2410–2422
Elgendy I, Muthanna A, Hammoudeh M, Shaiba HA, Unal D, Khayyat
proposed approach used hybridization-based consolida- M (2021) Security-aware data offloading and resource allocation
tion. During 100 tasks-based consolidation processes, our for MEC systems: a deep reinforcement learning. https://​doi.​org/​
proposed approach is taken a minimum time of 675 ms. 10.​36227/​techr​xiv.​13635​065.​v1
Elshabka MA, Hassan HA, Sheta WM, Harb HM (2020) Security-
aware dynamic VM consolidation. Egypt Inf J. https://​doi.​org/​
10.​1016/j.​eij.​2020.​10.​002
Gao Y, Guan H, Qi Z, Hou Y, Liu L (2013) A multi-objective ant
7 Conclusion colony system algorithm for virtual machine placement in cloud
computing. J Comput Syst Sci 79(8):1230–1242
Ghetas M (2021) A multi-objective Monarch Butterfly Algorithm
The energy-efficient virtual machine consolidation and for virtual machine placement in cloud computing. Neu-
migration based on adaptive beetle swarm optimization ral Comput Appl 33:11011–11025. https://​d oi.​o rg/​1 0.​1 007/​
(ABSO) has been explained in this paper. This method s00521-​020-​05559-2
effectively decreases energy consumption, cost, and mem- Hariharan B, Raj DP (2019) A hybrid framework for job scheduling
on cloud using firefly and BAT algorithm. Int J Bus Intell Data
ory utilization. For scheduling, ABSO has been explained. Min 15(4):388–407
The mathematical expression of PSO, BSO has been Hariharan B, Raj DP (2020) WBAT job scheduler: a multi-objective
clearly explained. Moreover, the MOF has been designed approach for job scheduling problem on cloud computing. J Circ
to decreases energy consumption and maximize sleeping Syst Comput 29(6):1–26
Huang, Qiang, Gao F, Wang R, Qi Z (2011) Power consumption of
PMs. The effectiveness of the recommended model has virtual machine live migration in clouds. In: 2011 third interna-
been discussed based on various metrics. The effectiveness tional conference on communications and mobile computing, pp
of the proposed VM consolidation and migration com- 122–125
pared with different methods and recommended model Khani H, Latifi A, Yazdani N, Mohammadi S (2015) Distributed con-
solidation of virtual machines for power efficiency in heterogene-
achieved the minimum values of memory, energy con- ous cloud data centers. Comput Electr Eng 47:173–185
sumption, and migration. The results obtained prove that Madhan ES, Srinivasan S (2014) Energy aware data center using
our recommended model effectively manages the resources dynamic consolidation techniques: a survey. In: 2014 proceedings
in the servers by minimizing the number of active servers of IEEE international conference on computer communication and
systems ICCCS14, pp 043–045
and ultimately minimizing the overall energy consump- Mohiuddin I, Almogren A (2019) Workload aware VM consolidation
tion. In the future, we will implement VM consolidation method in edge/cloud computing for IoT applications. J Parallel
adaptive selectors in real clouds such as OpenStack or Distrib Comput 123:204–214
CloudStack. And also, we will more concentrate on secu- Paulraj GJL, Francis SAJ, Peter JD, Jebadurai IJ (2018) Resource-
aware virtual machine migration in IoT cloud. Future Gener Com-
rity issues on migration. put Syst 85:173–183
Saeedi P, Shirvani MH (2021) An improved thermodynamic simu-
lated annealing-based approach for resource-skewness-aware and
Declarations power-efficient virtual machine consolidation in cloud datacent-
ers. Soft Comput 25(7):5233–5260
Saxena D, Singh AK, Buyya R (2021) OP-MLB: An online VM predic-
Conflict of interest statement The authors declares that they have no tion based multi-objective load balancing framework for resource
competing interest. management at cloud datacenter. IEEE Trans Cloud Comput.
https://​doi.​org/​10.​1109/​TCC.​2021.​30590​96
Sayadnavard MH, Haghighat AT, Rahmani AM (2021) A multi-objec-
References tive approach for energy-efficient and reliable dynamic VM con-
solidation in cloud data centers. Eng Sci Technol Int J. https://d​ oi.​
org/​10.​1016/j.​jestch.​2021.​04.​014
Ahmad RW, Gani A, Hamid SHA, Shiraz M, Yousafzai A, Xia F (2015) Sharifi M, Salimi H, Najafzadeh M (2012) Power-efficient distributed
A survey on virtual machine migration and server consolidation scheduling of virtual machines using workload-aware consolida-
frameworks for cloud data centers. J Netw Comput Appl 52:11–25 tion techniques. J Supercomput 61(1):46–66
Aryania A, Aghdasi HS, Khanli LM (2018) Energy-aware virtual Singh G, Singh AK (2021) Optimizing multi-VM migration by allocat-
machine consolidation algorithm based on ant colony system. J ing transfer and compression rate using geometric programming.
Grid Comput 16(3):477–491 Simul Model Pract Theory 106:102201
Beloglazov A, Buyya R (2012) Managing overloaded hosts for dynamic Suakanto S, Supangkat SH, Saragih R (2012) Performance measure-
consolidation of virtual machines in cloud data centers under ment of cloud computing services. Int J Cloud Comput: Serv Arch
quality of service constraints. IEEE Trans Parallel Distrib Syst 2(2):9–20
24(7):1366–1379

13
ABSO: an energy‑efficient multi‑objective VM consolidation using adaptive beetle swarm… 2197

Thakur S, Kalia A, Thakur J (2013) Server consolidation algorithms Zheng Q, Li R, Li X, Shah N, Zhang J, Tian F, Li J (2016) Virtual
for cloud computing environment: a review. Int J Adv Res Comput machine consolidated placement based on multi-objective
Sci Softw Eng 3(9):379–384 biogeography-based optimization. Future Gener Comput Syst
Witanto JN, Lim H, Atiquzzaman M (2018) Adaptive selection of 54:95–122
dynamic VM consolidation algorithm using neural network for
cloud resource management. Future Gener Comput Syst 87:35–42 Publisher's Note Springer Nature remains neutral with regard to
Yadav R, Zhang W, Kaiwartya O, Singh PR, Elgendy IA, Tian YC jurisdictional claims in published maps and institutional affiliations
(2018) Adaptive energy-aware algorithms for minimizing energy
consumption and SLA violation in cloud computing. IEEE
6:55923–55936
Zhang WZ, Elgendy IA, Hammad M, Iliyasu AM, Du X, Guizani M,
Abd El-Latif AA (2020) Secure and optimized load balancing for
multitier iot and edge-cloud computing systems. IEEE Internet
Things J 8(10):8119–8132

13