You are on page 1of 28

Wireless Personal Communications (2022) 124:75–102

https://doi.org/10.1007/s11277-021-09319-w

Secure VM Migration in Cloud: Multi‑Criteria Perspective


with Improved Optimization Model

Garima Verma1

Accepted: 24 October 2021 / Published online: 14 January 2022


© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2021

Abstract
Distributed computing has risen as a well-known worldview for facilitating an assortment
of online applications and services. The present business distributed computing stages uti-
lize a semi concentrated design, where cloud resources, such as servers and storage are
hosted in a few large global data centers. Virtualization in computing is a creation of vir-
tual (not real) of something such as hardware, software, platform or an operating system
or storage, or a network device. Further, Virtual Machine (VM) technology has recently
emerged as an essential building block for data centers and cluster systems, mainly due
to its capabilities of isolating, consolidating, and migrating workload. Migration of VM
seeks to improve the manageability, performance, and fault tolerance of systems. In a vir-
tual cloud computing environment, a set of submitted tasks from different users are sched-
uled on a set of Virtual Machines (VMs), and load balancing has become a critical issue for
achieving energy efficiency. Thus to solve this issue and to achieve a good load balance, a
new improved optimization algorithm is introduced namely Dual Conditional Moth Flame
Algorithm (DC-MFA) that takes into account of proposed multi-objective functions defin-
ing the multi-constraints like CPU utilization, energy consumption, security, make span,
migration cost, and resource cost. The performance of the proposed model will be analyzed
by determining migration cost, energy consumption, and response time, and security analy-
sis as well.

Keywords Cloud Computing · Virtualization · VM migration · Load balancing · Proposed


DC-MFA Model

Abbreviations
ABC + BA Artificial Bee Colony–Bat Algorithm
ACO Ant Colony Optimization
CSA Crow Search Algorithm
CSP Cloud Service Provider
DC-MFA Dual Conditional Moth Flame Algorithm
DFTM Dynamic Fault Tolerant VM Migration

* Garima Verma
garimaverma.research@gmail.com
1
Assistant Professor, School of Computing, DIT University, Vedanta Building,
Dehradun, Uttarakhand, India

13
Vol.:(0123456789)
76 G. Verma

ILWOA Improved Levy based Whale Optimization Algorithm


KH Krill Herd
MFO Moth Flame Optimization
OS Operating System
PM Physical Machine
TOA Time Of Appearance
VN Virtual Network
VMMAGS VM Migration Algorithm Based On Group Selection
VMM VM Monitor
VM Virtual Machine
WOGA Whale Optimization Genetic Algorithm

1 Introduction

Over the decades, cloud computing is one of the fastest-growing technology. NIST had
defined cloud computing as a “Model for enabling convenient, on-demand network access
to a shared pool of configurable computing resources (network devices, servers, storage,
applications, and services) that can be rapidly provisioned and released with minimal
management effort or service provider interaction”. Cloud computing has developed as a
mainstream worldview for facilitating an assortment of online applications and administra-
tions [1−4]. For instance, for safe storage and access, the medical apps make advantage of
Cloud Computing technologies. It is utilized to monitoring and detecting critical diseases
like cancer [5], cystic fibrosis [6], Brachymetatarsia [7] and so on to provide better services
to people through online healthcare applications. The present business distributed com-
puting stages utilize a semi concentrated design, where cloud assets, for example, servers
and capacity are facilitated in a couple of huge worldwide server farms. At the same time,
it gives economies of scale and factual multiplexing of assets over a huge client base. In
general, it, as a rule, suggests facilitating a cloud application at a cloud site that might be
removed” from its end clients. Distributed cloud computing empowers the client to get the
administrations on request over the web. The administrations offered are stage, program-
ming, and foundation. The clients pay just for the administrations they get to dependent on
their prerequisites. The specialist organizations guarantee a specific help level consent to
the customer. This empowers better asset provisioning for the customers.
The physical servers in the cloud are referred as a host and these hosts along with the
storages connected by network devices are referred as a data center. Each host or physical
machine in the cloud encompasses a huge count of VMs. Typically; the VM is the whole
OS of those acquaintances with applications and services. Furthermore, virtual machine
innovation has recently established as a fundamental structure hindrance for server farms
and group frameworks, due to its ability to disconnect, solidify, and relocate significant
strain [8–10]. Out and out, these highlights permit a server farm to serve different clients in
a safe, adaptable, and effective way. Therefore, these virtualized frameworks are thinking
about a key segment to drive the rising Cloud Computing worldview [11–14]. Relocation
of VMs looks to improve reasonability, execution, and adaptation to non-critical failure of
frameworks. All the more explicitly, the reasons that legitimize VM movement in a cre-
ation framework include the need to adjust framework load, which can be cultivated by
moving VMs out of over-burden/overheated servers [15–17].

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 77

A critical challenge in the cloud computing platform is the effective control of resources.
Cloud load balancing is a “state of the art” research area. Load balancing is the method of
reallocating the total load to each of the PM’s corresponding VMs to maximize resource
utilization while reducing response time [18]. Load balancing techniques are categorized
as static or dynamic depending on the state of the system. The dynamic method is supe-
rior to the static method since it does not require prior knowledge. In cloud computing,
load balancing is regarded as a multi-objective optimization challenge. Since load balanc-
ing must be carried out with sufficient resource utilization and without compromising the
user’s service quality [19]. It is important to choose the best physical host efficiently in the
process of deploying tasks to achieve better resource efficiency in the load balancing of
cloud data centers.
Traditionally, load balancing algorithms use a set of techniques to optimize candidate
target hosts during an algorithm period. Load balancing is done after selecting the best
target hosts. While the immediate effect can achieve high resource utilization, it does not
guarantee high execution performance for the next task [20]. This is greatly tolerated with
heuristics methodology, and some of the common approaches used are ABC, ACO [21],
etc. However, the scenario needs some promising approaches [21–26] for reducing the
energy consumption in VM migration.
The major contributions of this research work are:

• This research is presented for VM allocation of the user subjected to sixfold multi-
objective functions such as CPU utilization, energy consumption; make span, security,
migration cost, and resource cost.
• The proposed Dual Conditional Moth Flame Algorithm (DC –MFA) is an improved
version of the traditional Moth Flame Optimization (MFA).
• The rest of the paper is organized as: Section 2 addresses the recent works accom-
plished under VM migration. Section 3 depicts about proposed VM migration
approach: a compact architectural description. Section 4 tells about the problem defini-
tion and multi-objectives for quantifying VM migration. The resultants acquired with
the presented approach are discussed in Section 5 and a strong conclusion is provided
to the current research work in Section 6.

2 Literature Review

2.1 Related Works

In 2019, Sivagami and Easwarakumar [3] have proposed a novel DFTM to enhance the
reliability of the cloud data centre infrastructure. This was accomplished with the aid of a
sophisticated revival mechanism of demands of VN The authors have evaluated the route
traffic by means of implementing the Integer Linear Programming model, which takes into
deliberation the concern numerical factors that aid in choosing the optimal VM. The rout-
ing of new VM migration was undergone with the help of the Alternative Switch Identifi-
cation Algorithm. Finally, the resultant acquired from the planned fault recovery algorithm
had enhanced the VM survivability on the basis of the cloud data center.
In 2017, Guo et al. [27] have introduced VMMAGS with the intention of optimizing
the migration performance with respect to migration cost, communication cost, and VM

13
78 G. Verma

heat. The migration options were nothing but the selected VM groups and on the basis
of the integration cost of the chosen VM groups, the optimal migration scheme was
acquired. The resultant of the projected work had revealed a reduction in the migration
cost and communication cost.
In 2018, Karthikeyan et al. [28] have developed a new approach for VM migration
using the continuous Markov Chain. This approach enhances the security of the clouds
and prevents attack propagation. They have modeled the virtual machine behavior as
CTMC model by means of permitting the quantitative security assessment.
In 2020, Sutar et al. [29] have proposed a dynamic and energy-efficient live VM
migration approach that lessens the power wastages of inactive physical machines and
hence had resulted in savings of energy. The proposed model had enclosed seven mod-
ules: analysis on resource monitoring, Capacity distributor, Task allocator, Optimizer
analysis, Local Migration Agent, Migration Orchestrator, and Energy Manager. Further,
with the Ant Colony Optimization (ACO), the underloaded and overloaded physical
servers were identified. Further, the experimental evaluation had exhibited successful
VM migration with reduced energy consumption.
In 2019, Liu et al. [30] have introduced an adaptive method to achieve the schedul-
ing algorithms. Then, with the aid of Stochastic Reward Nets, they have modeled the
migration and placement process of VMs. The adaptive evaluation was performed with
the reward functions acquired from diverse scheduling approaches. The proposed model
was evaluated in terms of quality of service and the cost of runtime.
In 2018, Narayanan and Saravanaguru [31] have developed a novel approach for
securing IP Sec tunneling during live VM migration from one physical infrastructure to
another. To avoid the man in the middle attack, they have encrypted the router to router
path. Since they had encrypted the channel, the security of the network was higher.
Further, they had enhanced the security of the VM migration with an onion routing
algorithm.
In 2019, Torquato et al. [32] have introduced a Continuous Time Markov Chain model
for choosing the satisfactory server for migration. This arrangement had permitted to per-
form quantitative security appraisal by displaying the conduct of the VM as CTMC model.
The major goal of CTMC was to assess the likelihood of a VM to be undermined depend-
ent on the goal server parameters and locate the best and fitting resistance to diminish
the effect of assaults. Various situations was created to represent quantitatively in which
case the VM’s movement was undermined and to examine the adequacy of the proposed
arrangement in term of physical attack.
In 2019, Kandoussi et al. [33] have built up a security assessment approach depend-
ent on an accessibility model for virtualized frameworks with VM relocation for VMM
restoration. The objective was to locate the best possible revival planning to arrive at the
ideal levels (or if nothing else to keep away from the undesired degrees) of security hazard
and accessibility. They have presented three contextual analyses involving significant secu-
rity dangers, to be specific Man-in-the-middle and Denial of Service assaults. Results give
canny data with respect to the trade-off among accessibility and security hazards while
applying VM relocation planning for revival purposes.
In 2018, Anurag satpathy et al. [34], have introduced a two-phase VM placement tech-
nique. Initially, a queuing framework was presented to arrange and direct the great amount
of VM. After that, CSA-based VM placement was utilized to minimize the resource dev-
astation and power consumption at servers. This CSAVMP algorithm selects a server for
placement from the available servers once a VM was scheduled for placement.

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 79

In 2019, soltanshahi et al. [35], have suggested an optimized VM placement with the
aid of the KH algorithm. As a response, this method was implemented for reducing energy
consumption by optimizing virtual machine aggregation and closing down idle servers
while maintaining service quality. The simulation results show that a well-integrated sys-
tem and the use of practical VM migration strategies will help to boost energy efficiency.
In 2021, Saxena et al. [36] have introduced a secure and multi-objective VM placement.
In this, whale evolutionary algorithm and non-dominated sorting-based genetic algorithm
were combined and Whale Optimization Genetic Algorithm (WOGA) was proposed for
VM placement. By lowering inter-communication delay, the introduced system enables an
energy-efficient distribution of physical resources among VMs, emphasizing the secure and
timely execution of user applications.
In 2019, Mohamed et al. [37] have suggested a levy-based whale optimization algo-
rithm. The authors devised a cost-effective VM placement mechanism that saves energy
and reduces resource waste. The work done has only one goal: to reduce the utilization
rate of PMs in cloud data centers. To fit the most VMs into the smallest number of PMs,
a variable-sized bin packing issue is used. Furthermore, the best-fit approach is applied to
provide the greatest results.

2.2 Review

The most important literary works are discussed in the literature section and their features
are exhibited in Table 1. Some disadvantages in existing methods are resource utiliza-
tion decreases at the duration of increased migration time [3], overload of network time
[28], high migration time [29], large computation time [30], large variation in response
time [31], cannot adapt in the complex network [33], time-consuming [34], slow conver-
gence [35], and so on. To, overcome such issues we proposed a novel secure VM migra-
tion approach with efficient load balancing. Comparing to existing methods, our model was
enhanced based on a sixfold multi-objective function. Result analysis shows that how our
method is efficient than other methods.

3 Proposed VM Migration Approach: A Compact Architectural


Description

To hold high performance and availability, VMs needs to be migrated from one cloud to
every other to leverage higher resource availability, avoid downtime due to hardware main-
tenance, and shop greater energy within the source cloud. In cloud computing environ-
ment, the importance of load balancing has not to be overstated. The most efficient method
of load balancing is to migrate overloaded VMs to underloaded VMs with optimized
resource utilization. The objective of VMs migration is to limit the assignment execution
time and maximize the assets utilization via load balancing within the dynamic environ-
ment. A massive quantity of studies efforts has been dedicated to investigating the chal-
lenge VM migration in cloud systems over the final decade.

3.1 Proposed Architecture: Cloud Platform Model and Task Requirement

Figure 1 explains the proposed model of secured VM migration on efficient load balancing.

13
80

Table 1  Features and challenges of existing VM migration approaches

13
Author [Citation] Methodology Features Challenges

Sivagami and Easwarakumar [3] DFTM algorithm Proficiently increasing the cloud survivability with Resources should be utilized properly VM increases, the
minimal complexity Takes very little time for VM resource wastage also increases
migration
Guo et al. [27] VMMAGS Reduce the migration cost and communication cost Response time fluctuated greatly during migration Prone
Improve the system reliability to ineffective migration
Karthikeyan et al. [28] ABC–BA Able to achieve the minimum energy consumption and Overload of network traffic requires efficient utilization of
failure rate Limits the quantity of failure in VM migra- cloud resources
tion occasions
Sutar et al. [29] ant colony optimiza- Provides minimum migration time Reduces the energy Need to reduce the size of VM before migration Need to
tion algorithm consumption further reduce the migration time
Liu et al. [30] Stochastic Petri Nets High throughput Lower is the mean service time Low availability job service rate is low Power consump-
tion needs to be reduced
Cao et al. [31] HMGOWM Save the utilization of the CPU Reduce the migration Need to avoid the disallowed overlap
probability of VMs with high access Probability
Fu et al. [32] LVMM Can reduce the migration cost Improve the efficiency of Great extra cost
migrations
Yang et al. [33] SnapMig Redundant data blocks are identified and removed Extra hardware cost. Degrades both IO throughput and
Shorter migration time migration time
Anurag satpathy et al. [34] CSA Optimize the resource wastage and power consumption Time-consuming
Soltanshahi et al. [35] KH Reduce energy consumption in data centers No guarantee for fast convergence solution
Saxena et al. [36] WOGA Low power consumption, resource wastage, and com- The mapping of virtual machines to various server clus-
munication cost ters is not taken into account
Mohamed et. al [37] ILWOA Efficient bandwidth Reduce the count of running PM Need to consider the multi-objective VM placement and
dynamic migration
G. Verma
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 81

Fig. 1  Block diagram of proposed VM migration for efficient load balancing

In the cloud platform framework, a resource pool comprises an arrangement of serv-


ers, every server is represented as Si, and server resources group of the cloud condition is
Si ;i = 1, 2, ..., n. Under the beginning state k , VMs are accepted conveyed on the server Si
at every node and it is denoted as Si = VMi,1 , ..., VMi,k. Here, VMi,k is a VM on the server Si.
Every server Si contains numerous equipment assets, for example, Network, CPU, memory.
In cloud conditions, the dynamic determination of host for VM allotment is a significant task.
The client gives a server demand for service as {Ti ;i = 1, 2, .., n} and here each sort
of administration has QoS necessity. The solicitations for a similar kind of administra-
tion may have unique prerequisites for TOA, fruition time, and so on. The client pre-
sents the undertaking Ti ;i = 1, 2, .., n to VMs in the cloud platform. VM groups will
improve task planning as indicated by task prerequisite and accessibility assessment of
VM resources. Each task of Ti ;i = 1, 2, .., n is characterized( by a )sixfold of parameters:
Ti = {UCPU , En, Sec , Ms, CoM, Res}. Here, CPU utilization UCPU , energy consumption
(En), security (Sec), make span (Ms), migration cost (CoM), and resource cost (Res) are the

13
82 G. Verma

key terms for efficient load balancing during VM migration. This is considered to be an opti-
mization problem and a new algorithm is introduced for resolving it.

4 Problem Definition and Multi‑Objectives for Quantifying VM


migration

4.1 Multi‑objective Function: Description

The major objective of this work relies upon the minimization of the sixfold multi-objec-
tives, which is mathematically defined in Eq. (1).
{ }
obj = min UCPU , En, Sec , Ms, CoM, Res (1)

4.1.1 CPU Utilization

UCPU Typically, CPU utilization is the sum of time for which the CPU is utilized to process
the specified task [38]. Also, when an application demands for an asset on the cloud, VMs
are mapped with pools of the physical server. These VMs are so set, to satisfy the CPU use
of its host with the goal that various undertakings should be possible without a moment’s
delay. It is computed with respect to the tasks as per Eq. (2).

UCPU (tasks) = CountCPU(Tq +,...+Tn ) (2)

4.1.2 Energy Consumption

The energy consumption shows a straight relationship with the usage of PM approximation
[39]. “The energy consumption at a data center is defined as a total amount of power (P)
consumed over a period of time while performing the work”. It is computed using Eq. (3).
Here, Eni overall energy consumed at the time, when all VM are active and CountCPU is the
count of CPU’s.
( )
En = 0.1 log CountCPU + Eni + 0.1 (3)

4.1.3 Security

This migration of VM also needs to be secure from theft for efficient and secured load
balancing. The security of VM (X(i)) is computed by evaluating the difference between the
security of the physical machine
( and ) the task security. This evaluation is undergone with
the aid of the risk probability Prisk model. The pseudocode is shown in Algorithm 1

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 83

4.1.4 Make Span

A focal issue in the planning hypothesis is to structure a schedule to such an extent that the last
completing time of the given occupations (likewise called makespan) is limited. The makes-
pan of an undertaking is the separation in time that passes(from )the beginning of work as far as
possible. This is computed to be the maximum wall time Wtime of overall blocks in the tasks.
(N )

Ms(tasks) = max Wtime (4)
B=1

4.1.5 Migration Cost (CoM)

The CoM is likewise a significant measurement for assessing the scheduling algorithm. The
CoM can be processed by the recurrence of relocation execution, which is related to the
accompanying prize capacity. It is the sum of the cost incurred by VM during its migration
from one task to other. It is mathematically defined in Eq. (5)
n
∑ ( )
CoM = CoM Ti;i=1,2,..,n (5)
Ti =1

13
84 G. Verma

4.1.6 Resource Cost

The expense of relocation relies upon numerous elements including (1) the memory
content and the memory update pace of each VM, (2) the all out number of VMs to be
relocated, (3) the accessible system data transfer capacity for movement, (4) the remain-
ing burden of the source and the goal servers at the hour of relocation. From the spe-
cialist co-op purpose of viewpoint, this component empowers to powerfully extend and
recoil resource requests, in a perfect world fitting the expense of rented resources to the
foreseen remaining task at hand. So also, the foundation supplier can maintain a strate-
gic distance from the wasteful utilization of resources, because inactive or underutilized
resources can be turned off. The remaining task at hand solidification (or server union)
is one of the components to productively use resources in distributed computing. In this
situation, the foundation supplier normally screens the size of the approaching remain-
ing task at hand and the dispersion of the remaining task at hand inside the cloud condi-
tion to decide the quantity of servers it should make accessible. ( )
Mathematically (in Eq. 6), resource cost is the sum of the cost ResCost incurred by
the resources for performing the corresponding task.

Res(tasks) = ResCost(Tq +...+Tn ) (6)

5 Proposed Optimization Approach for VM migration with Solution


Encoding

5.1 Solution Encoding

The solution encoding is demonstrated in Fig. 2. The chromosome length is twice the
multiple of the count of blocks Bin ;i = 1...N allocated for overall tasks. Each of the
blocks encompasses two parts: the PM, which performs the task and the id of VM
(VM id ) to whom the task is allocated by PM. Typically, the block size varies for each of
the tasks.
For illustration: if the count of blocks = 42, then the chromosome length = 84. Here,
the 1st 42 chromosomes are filled by PM and the rest of them are filled by VM.

Fig. 2  Solution Encoding

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 85

5.2 Proposed DC‑MFA Optimization

One extravagant creepy-crawly that is similar to the group of butterflies is the moth [22] [40]
[41]. The transverse direction component ensures the straight-line movement of the moth
since the moon is far away from it. The means followed in the proposed DC-MFA improve-
ment are portrayed underneath:
Step 1: The overall population of the moth Mothm and the flame Flamew are initial-
ized. Here, Mothm is the mth moth of wth flames. The overall count of flames is Countflames
and q is the current count of iteration. In addition, a random variable rand1 and rand2 is also
initialized.

• Fitness Function for all moths is calculated by using the Eq. (1).

Step 2: When q <= Countflames.


(a) If rand1 < 0.5
The position of the moth, as well as flame, is updated using the standard MFO algorithm.
The steps followed are described in the subsequent section:

• The transverse orientation mechanism for the positional update of the moth with respect to
flame is mathematically expressed in Eq. (7).
( )
Mothm = S Mothm , Flamew (7)

• The logarithmic spiral position update of the search agent is mathematically expressed in
Eq. (8).
( )
S Mothm , Flamew = Rm .ebt . cos(2𝜋t) = Flamew (8)

Here, Rm defines the shape of the logarithmic spiral for the wth flame of mth moth and it is
mathematically expressed in Eq. (9)

Rm = ||Flamew − Mothm || (9)

b Constant (random number).


t Lies within [ − 1, 1].

• Further, over the course of iterations t , the count of flames can be computed as per
Eq. (10).
( )
N−1
Countflames = round N − q × (10)
T
Here, N and T , the maximum count of flames and maximum count of iterations,
respectively.
(b) If rand > 0.5

• Compute the distance between to flame dis from moth using Eq. (11).
dis = best pos(Flame) − mothpos (11)

mothpos = dis + best pos(Flame) (12)

13
86 G. Verma

Step 3: Further, When q > Countflames.


(a) If rand < 0.5
The position of the moth, as well as flame, is updated using the standard MFO algo-
rithm, which is mathematically defined above (in Eq. (10) to Eq. (13)).
(b) If rand2 > 0.5
The position update of the search agent (moth and flame) takes place using the newly
proposed mechanism.
Compute the distance between to flame dis from moth using Eq. (13) and Eq. (14),
respectively.
dis = best pos(Flame) − sorted (pop) (13)

mothpos = dis + best pos(Flame) (14)


The pseudo-code of the proposed model is portrayed in Algorithm 2.

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 87

5.3 Database Creation

The database collected for VM migration- load balancing is partially synthetic, i.e. the
data corresponding to the physical machine are downloaded from: https://​www.​kaggle.​
com/​discd​iver/​clouds [Access date: 2020–04-26]. Using this database, the corresponding
data related to VM are generated. The database of PM encompasses certain parameters are
shown in Table 2.
The database for VM is created with the aid of the PM database. It also encompasses
the same parameters as the PM. The database creation mechanism for each of the con-
straints is discussed below:

13
88 G. Verma

Table 2  Database constraints 1 Count of PM Original data

2 Number of tasks Synthetic data


3 Wall time
4 Cost of PM
5 Security
6 Count of CPU

Let the task to be accomplished in PM be Ti ;i = 1, 2, 3, 4. The block size of task 1, task


2, task 3, and task 4 is 5, 10, 15, and 12, respectively. The task from the user is allocated
to PM, which in terms splits the tasks and schedules it in a separate VM. Here, 13 PM are
utilized and each PM is said to have 10 VM’s. So, there is a total of 130 VM’s to perform
the tasks.

5.3.1 Security

PM has its security range fixed randomly between 1 to 4. For illustration, if the count of
tasks to be accomplished = 6, then, 6 counts of security values will be created within the
bounds (1–4). Here, the task with security value 1 is said to be highly secured, while the
least secured task takes the highest security value (4).
The security being the important parameter, it is computed for VM by evaluating the
difference between the PM security and task security. As mentioned earlier, the security
value of PM and VM is fixed between 1 and 4. For illustration, if the security value of PM
is 1 and task security is 1, then the security of VM is 0. Thus, it is said to have fulfilled the
required security, as per the risk mechanism.

5.3.2 Wall time

“Elapsed real-time, real-time, wall-clock time, or wall time is the actual time taken from
the start of a computer program to the end. In other words, it is the difference between the
time at which a task finishes and the time at which the task started. Wall time is thus differ-
ent from CPU time, which measures only the time during which the processor is actively
working on a certain task. The difference between the two can arise from architecture and
run-time dependent factors, e.g. programmed delays or waiting for system resources to
become available”. The PM has a pre-defined wall time. The wall time of VM is computed
using the “normrnd function”. The syntax of normrnd function expressed in Eq. (15)
Wtime = normrnd(𝜇, 𝛿) (15)
This function typically creates a random number from the normal distribution with
mean parameter 𝜇 and standard deviation parameter 𝛿 sigma. Here, 𝛿 is fixed as 0.1 and 𝜇 is
the wall time of PM.

5.3.3 VM cost

The cost of VM is also computed with the “normrnd function”. Here, 𝛿 is fixed as 0.1 and
𝜇 is the cost of PM.

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 89

VMcost = normrnd(𝜇,𝛿) (16)

5.3.4 Migration cost

The migration cost of VM is computed from the M × M matrix, whose column as well as
rows point out to the migration cost of PM. Here, the matrix with equivalent rows and col-
umns elements, i.e. (1,1), (2,2) as well shows reduced migration cost, while the others have
higher migration cost.

5.3.5 Energy consumption

The energy consumed by VM is also computed using the “normrnd function”.


En = normrnd(𝜇,𝛿) (17)
Here, the overall power consumed by PM is taken as 63.40 W and it is the mean 𝜇 of
normrnd . The standard deviation 𝛿 is generated with a random number ±5%. In case if the
task in block 1 is being accomplished in 4th VM of 2nd PM, then the CPU utilization of
block 1 is 2 + 4 = 6. Alike this, the CPU utilization of other blocks is added and finally, the
overall CPU utilization count is computed for all the tasks by summing them all together.

6 Results and Discussions

6.1 Simulation Procedure

The proposed VM migration approach for efficient load balancing with optimization algo-
rithm was carried out in MATLAB and the experimented outcome was investigated. The
performance of the proposed model has been analyzed by determining migration cost, total
cost, energy consumption, response time, and security analysis as well. The presented work
(DC-MFA) was compared over the existing works like ABC + BA [28], ACO [29], CSA
[34], KH [35], WOGA [36], and ILWOA [37] respectively. The evaluation is done by vary-
ing the VM and blocks. The VM is varied from [10 20 30 40 50] for a PM and Block vari-
ation is from [20, 25, 30, 35, 40] (in graphical results). Further, sub-blocks will be created
from the above block variation and here T1, T2, T3, T4 are tasks.
As per the collected database.

• Block variation 1 encompasses: [4 7 6 3]-[T1 T2 T3 T4] = total blocks 20


• Block variation 2 has [6 8 6 5]-[T1 T2 T3 T4] = total blocks 25
• Block variation 3 has [6 10 7 7]-[T1 T2 T3 T4] = total blocks 30
• Block variation 4 has [7 11 8 9]-[T1 T2 T3 T4] = total blocks 35
• Block variation 5 [5 10 15 10]-[T1 T2 T3 T4] = total blocks 40

6.2 Evaluation on CPU Utilization

CPU utilization is the sum of time for which the CPU is utilized to process the guid-
ance of a program. Correspondingly, when an application demands for an asset on a

13
90 G. Verma

cloud, VMs are mapped with pools of a physical server. These VMs are so set, to sat-
isfy the CPU usage of its host with the goal that various tasks should be made possible
immediately. Therefore, it is essential to have lower CPU utilization. The resultant of
CPU utilization for the presented work over the existing works is graphically shown
in Fig. 3. On observing, the presented work takes less CPU utilization even in all vari-
ations (count of blocks) in the tasks. Particularly, for 10 counts of VMs in a PM as
per Fig. 3a, the presented work is 40%, 43.40%, 25%, 72.73%, 45.45%, and 50% better
than the traditional approaches like ABC + BA, ACO, CSA, KH, WOGA, and ILWOA
respectively for 20 counts of blocks. The overall evaluation of CPU utilization for each
VM by proposed over existing is tabulated in Table 3. As per the objective, the CPU uti-
lization is lower in the presented work, while compared to the existing works. When 30
VM is allocated to a PM for doing a certain task, the proposed model achieves the CPU
utilization as 965.12, while the existing works have ABC + BA = 1598.5, ACO = 1095.6,
CSA = 1254.2, KH = 2670.8, WOGA = 3219.6, and ILWOA = 2213.8. Altogether, the
proposed work proves its betterment with respect to minimum CPU utilization.

Fig. 3  Evaluation on CPU utilization for presented work over existing work in case of a varying count of
VMs allocated by a PM for load balancing a VM = 10 b VM = 20, c VM = 30 and VM = 40

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 91

Table 3  Overall CPU utilization for diverse VM Allocated by a PM for exiting and proposed model
Count of VM ABC + BA[28] ACO[29] CSA[34] KH[35] WOGA [36] ILWOA[37] DC-MFA

10 935.61 1323.2 1253.9 1280.8 4712.1 5692.2 886.2


20 1679.2 1535.9 1202.9 1790 3290.9 5071.9 1228.8
30 1598.5 1095.6 1254.2 2670.8 3219.6 2213.8 965.12
40 1598.9 1614.7 1501.7 1675.3 1858.5 3121.6 1135
50 985.04 1311.8 1139.7 2246.8 2418.1 5818.5 1389.7

6.3 Evaluation on Energy Consumption

The energy consumption for the presented work over existing work for varying counts of
VM and blocks are graphically shown in Fig. 4. In Fig. 4c, the presented work has the low-
est energy consumption, while compared to the existing one, and for 30 counts of blocks,
the presented work is 1.62%, 4.21%, 6.67%, 5.21%, 2.67%, and 3.19% better than the
extant works like ABC + BA, ACO, CSA, KH, WOGA, and ILWOA respectively. Then, in
case of overall energy consumption in Table 4, the presented work consumes less energy
under a certain allocation of VMs for a PM. Even though the energy consumption remains

Fig. 4  Evaluation on Energy consumption for presented work over existing work in case of a varying count
of VMs allocated by a PM for load balancing a VM = 10 b VM = 20, c VM = 30 and VM = 40

13
92 G. Verma

Table 4  Overall performance: energy consumption for diverse VM Allocated by a PM for exiting and pro-
posed model
Count of VM ABC + BA[28] ACO[29] CSA[34] KH[35] WOGA [36] ILWOA[37] DC-MFA

10 2491.1 2530.7 2526.9 2558.2 2487.5 2553.7 2461.5


20 2491.6 2580.2 2566.4 2563 2537.6 2548.9 2588.7
30 2468.2 2560.2 2564.8 2580.5 2593.3 2585.8 2502.9
40 2548.3 2526.6 2544.8 2518 2524 2544.8 2623.3
50 2515.1 2531.2 2526 2549.6 2676.8 2562.3 2610

higher in some variations, the overall performance of the proposed work is satisfactory
when compared to other methods.

6.4 Evaluation on MakeSpan

The make-span is characterized as the measure of time, and through for finishing a lot of
occupations, for example, the most extreme finishing time all things considered. It should
be minimized for efficient load balancing in VM migration. The resultant of make-span is
exhibited in Fig. 5 for diverse VMs to one PM. In Fig. 5d, for 40 counts of VM at one PM

Fig. 5  Evaluation on make-span for presented work over existing work in case of a varying count of VMs
allocated by a PM for load balancing (a) VM = 10 (b)

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 93

corresponding to 35 counts of nodes, the presented work is 2.33%, 30%, 42.47%, 46.84%,
31.15%, and 33.33% better than the extant approaches like ABC + BA, ACO, CSA, KH,
WOGA, and ILWOA respectively. As a whole, the presented work has the lowest make-
span. The overall resultant of make-span values acquired for the presented over the extant
approaches is tabulated in Table 5. The overall makes-span of the presented work is the
lowest under a certain allocation of VMs to one PM. Further, while allocating 30 VMs to
one PM, the presented work has the lowest make-span as 281.55, which is 23.39%, 8.79%,
20.90%, 16.61%, 53.03%, and 39.73% better than existing ABC + BA, ACO, CSA, KH,
WOGA, and ILWOA respectively. Thus, as a whole, the overall objective is achieved by
the presented work.

6.5 Evaluation on Migration Cost

The migration cost of the presented work over the existing work is shown graphically in
Fig. 6. On observing all the variations in VM as well as block size, the presented work
has the lowest migration cost and hence satisfies the objective. At VM = 10, the presented
work is 64.81%, 61.22%, 59.57%, 60.42%, 65.45%, and 52.5% better than the existing
approaches like ABC + BA, ACO, CSA, KH, WOGA, and ILWOA for 40 counts of sub-
blocks. Thus, as a whole, the presented work has the lowest migration cost while compared
over the existing. The migration cost for the overall migration cost acquired for diverse
VM is shown in Table 6. The overall migration cost is lowest in the presented work under
every variation in the allocation of VM to a PM. Typically, while allocating 10 VMs to
one PM, the presented work has the lowest migration cost of 88.809, while the traditional
approaches have higher values as ABC + BA = 295.65, ACO = 259.1, CSA = 252.03,
KH = 291.75, WOGA = 239.93, and ILWOA = 287.19. Similarly, for allocating 40 VMs to
one PM, the proposed model is 33.62%, 44.49%, 47.92%, 47.54%, 37.17%, and 45.53%
better than existing approaches like ABC + BA, ACO, CSA, KH, WOGA, and ILWOA
respectively. As a whole, the presented work is said to achieve the objective of migration
cost minimization.

6.6 Evaluation on Security

Figure 7 represents the performance of the proposed work over other models with
respect to security. From the graphical representation, it is proved that the proposed
work ensures high security in load balancing with less risk. However, the conventional
models portray high risk that shows the unsecured scenario. particularly, the presented
work has tolerated with less risk at 35 counts of nodes in ­30th VM and it is 83.33%,

Table 5  Overall Performance: Make-Span for diverse VM Allocated by a PM for exiting and proposed
model
Count of VM ABC + BA[28] ACO[29] CSA[34] KH[35] WOGA [36] ILWOA[37] DC-MFA

10 285.62 371.6 293.18 306.6 525.19 308.16 284.77


20 329.3 333.89 374.82 318.85 398.76 391.87 357.76
30 367.49 308.7 355.96 337.64 599.45 467.14 281.55
40 421.67 283.27 277.17 360.91 542.41 375.02 302.27
50 322.62 384.46 347.43 391.23 605.21 331.45 379.59

13
94 G. Verma

Fig. 6  Evaluation on migration cost for presented work over existing work in case of a varying count of
VMs allocated by a PM for load balancing a VM = 10 b VM = 20, c VM = 30 and VM = 40

Table 6  Overall Performance: Migration Cost for diverse VM Allocated by a PM for exiting and proposed
model
Count of VM ABC + BA[28] ACO[29] CSA[34] KH[35] WOGA [36] ILWOA[37] DC-MFA

10 295.65 259.1 252.03 291.75 234.93 287.19 88.809


20 272.99 247.12 254 268.06 256.54 268.47 190.82
30 227.09 247.69 243.15 248.36 69.624 237.11 221.24
40 231.2 276.49 294.68 292.56 244.26 281.76 153.47
50 217.64 208.99 301.89 257.07 74.481 281.47 147.57

82.76%, 82.46%, 82.14%, 76.19%, and 87.80% better than ABC + BA, ACO, CSA,
KH, WOGA, and ILWOA respectively. The overall performance of the proposed work
under the security factor is tabulated in Table 7. However, some methods have good
security than the proposed model. But, overall or model provides better performance in
terms of security.

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 95

Fig. 7  Evaluation on security for presented work over existing work in case of a varying count of VMs allo-
cated by a PM for load balancing a VM = 10 b VM = 20, c VM = 30 and VM = 40

Table 7  Overall Performance: Security for diverse VM Allocated by a PM for exiting and proposed model
Count of VM ABC + BA[28] ACO[29] CSA[34] KH[35] WOGA [36] ILWOA[37] DC-MFA

10 0.088531 0.068857 0.11804 0.12212 0.50062 0.28106 0.22457


20 0.092612 0.082776 0.078694 0.13604 0.18888 0.18888 0.27208
30 0.16555 0.10245 0.068857 0.16104 0.60423 0.15571 0.2049
40 0.1332 0.19222 0.092612 0.17131 0.36543 0.21022 0.25816
50 0.088531 0.2229 0.1082 0.12869 0.58498 0.3159 0.32367

6.7 Evaluation on Resource Cost

The analysis on resource cost is exhibited in Fig. 8 and here the evaluation is undergone
for diverse VM by varying the count of blocks. The presented works show the lowest cost
for every variation in block count. Further, on observing, 1­ 0th VM for 40 counts of a block,
the presented work is 34.92%, 30.51%, 59.80%, 48.75%, 69.85%, and 66.12% better than
the existing works like ABC + BA, ACO, CSA, KH, WOGA, and ILWOA respectively. As
a whole, the lowest resource cost is consumed by the presented work. Table 8 shows the
overall resource cost allocated to a PM having a different count of VMs. The presented

13
96 G. Verma

Fig. 8  Evaluation on Resource cost for presented work over existing work in case of a varying count of
VMs allocated by a PM for load balancing a VM = 10 b VM = 20, c VM = 30 and VM = 40

Table 8  Overall Performance: resource Cost for diverse VM Allocated by a PM for exiting and proposed
model
Count of VM ABC + BA[28] ACO[29] CSA[34] KH[35] WOGA [36] ILWOA[37] DC-MFA

10 52.205 31.762 52.205 42.035 56.964 62.251 21.217


20 30.668 34.147 30.668 48.102 43.797 50.233 27.804
30 29.814 29.482 29.814 46.401 55.254 24.442 24.335
40 28.979 49.388 28.979 46.908 34.699 48.693 23.948
50 34.257 45.57 34.257 32.418 41.772 67.33 25.471

work has the lowest resource cost under every variation in the VM count. Further, to reveal
this achievement in minimization of resource cost in the presented work, the observation
is accomplished with 10 counts of VM a PM. Here, the presented work has the lowest
value as 21.217, which is 59.36%, 33.2%, 59.36%, 49.53%, 62.75%, and 65.92% better than
existing works like ABC + BA, ACO, CSA, KH, WOGA, and ILWOA respectively. For
40 counts of PM, the proposed model is 25.65%, 44.11%, 25.65%, 21.43%, 39.02%, and
62.17% superior to exiting ABC + BA, ACO, CSA, KH, WOGA, and ILWOA methods

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 97

respectively Thus the resultants proved that the proposed model provides better resource
cost over existing methods.

6.8 Evaluation on Total Cost

The total cost (overall objective) inclusive of migration cost, CPU utilization cost, resource
cost as well needs to be lower for efficient and secured load balancing. The resultant is
exhibited for diverse VM in Fig. 9. On observing, 30 count of VM, the presented work
has the lowest value and it is 5.05%, 8.29%, 8.29%, 29.06%, 10.48%, and 6% better than
the traditional works like ABC + BA, ACO, CSA, KH, WOGA, and ILWOA respectively
for 20 counts of nodes. The overall total cost acquired in terms of the total cost is tabu-
lated in Table 9. The overall cost of the presented work is the lowest one for every varia-
tion in the count of VM allocated to a PM. For 20 counts of VM, the proposed model is
8.75%, 7.12%, 0.78%, 11.91%, 32.68%, and 47.26% better than existing ABC + BA, ACO,
CSA, KH, WOGA, and ILWOA methods respectively. Likewise, for 40 counts of VM is
allocated for one PM, the proposed DC-MFA model is 12.25%, 10.78%, 8.80%, 13.39%,
18.56%, and 33.49% better than existing methods ABC + BA, ACO, CSA, KH, WOGA,
and ILWOA. Even though there might be smaller fluctuations in terms of individual objec-
tive functions, the overall objective is smaller for the presented work and hence said to be
much sufficient for load balancing in VM migration.

Fig. 9  Evaluation on the total cost for presented work over existing work in case of a varying count of VMs
allocated by a PM for load balancing a VM = 10 b VM = 20, c VM = 30 and VM = 40

13
98 G. Verma

Table 9  Overall performance: evaluation on total cost of proposed over existing models
Count of VM ABC + BA[28] ACO[29] CSA[34] KH[35] WOGA [36] ILWOA[37] DC-MFA

10 4044 4516.4 4378.3 4479.5 8017.2 8903.8 3742.8


20 4815.3 4731.3 4428.9 4988.2 6527.8 8331.6 4394.2
30 4708.1 4241.7 4448.1 5883.8 6537.8 5528.4 3995.3
40 4830.2 4750.6 4647.4 4893.8 5204.2 6372 4238.3
50 4068.2 4482.2 4349.3 5477.2 5816.9 9061.4 4552.6

6.9 Statistical Analysis: Proposed versus Conventional Models

To ensure a fair comparison, each algorithm is executed ten times to obtain the statistics
of the costs to be minimized. Table 10 shows the statistical resultant for varying counts

Table 10  Statistical evaluation for a varying count of VMs allocated by a PM for load balancing
VM = 10
Measures ABC + BA[28] ACO[29] CSA[34] KH[35] WOGA [36] ILWOA[37] DC-MFA

Best 1997.3 2043.5 1986.3 2755.1 2111.5 2111.7 1787.4


Worst 3735 3686 4314.8 4983.4 3808 3608.4 3422.4
Mean 2847.4 2909.6 3070.4 3564.7 2936.1 2951.6 2551.7
Median 2828.6 2954.4 2990.4 3260.1 2912.4 3043.2 2498.5
STD 749.99 699.48 990.64 1031.7 747.18 733.15 716.48
VM = 20
Measures ABC + BA[28] ACO[29] CSA[34] KH[35] WOGA [36] ILWOA[37] DC-MFA
Best 2049.6 2172.9 2206.1 2747.8 2124.1 2194.3 1983.4
Worst 3469.4 3613.8 3734.5 3799.8 4013.3 3657.7 3640.5
Mean 2764 2836.7 2970 3269.1 3021.4 2894.7 2751.5
Median 2768.6 2780.1 2969.6 3264.4 2974.2 2863.4 2691.1
STD 612.02 625.92 651.1 432.17 786.28 656.5 735.3
VM = 30
Measures ABC + BA[28] ACO[29] CSA[34] KH[35] WOGA [36] ILWOA[37] DC-MFA
Best 1947.5 2088.6 2093.3 2539.5 2169.6 2004.4 1686.1
Worst 3470.5 3649.9 3842.8 4102.8 4236.3 3626.8 3209.2
Mean 2698.6 2993 2980.7 3395 3055 2899.2 2540.3
Median 2688.2 3116.7 2993.3 3468.8 2907 2982.9 2633
STD 649.34 758.58 771.09 696.84 902.66 714.6 695.09
VM = 40
Measures ABC + BA[28] ACO[29] CSA[34] KH[35] WOGA [36] ILWOA[37] DC-MFA
Best 1914.5 1986.7 2047.8 2167.4 2356.4 2201.1 1764.9
Worst 3601.9 3608.8 3834.1 4692.1 4449.4 3891.3 3451.1
Mean 2749.5 2782.9 2955.5 3575.6 3204.8 2990.8 2662.8
Median 2740.8 2768 2969.9 3721.5 3006.6 2935.5 2717.7
STD 734.28 702.04 756.04 1105.3 943.95 761.01 701.33

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 99

of VMs allocated to one PM for performing the task scheduling function. On observing
the mean value for 10 counts of VMs, the presented work has the best lowest value as
1787.4, whereas, the existing models have higher mean values as ABC + BA = 1997.3,
ACO = 2043.5, CSA = 1986.3, KH = 2755.1, WOGA = 2111.5, and ILWOA = 2111.7.
When 20 counts of VM is allocated to one PM, the best value of the presented model is
3.23%, 8.72%, 10.09%, 27.82%, 6.62%, and 9.61% better than existing ABC + BA, ACO,
CSA, KH, WOGA, and ILWOA methods. Furthermore, in median case scenario, the pro-
posed DC-MFA model is 0.84%, 1.82%, 8.49%, 26.97%, 9.61%, and 7.42% superior to
existing ABC + BA, ACO, CSA, KH, WOGA, and ILWOA methods for 40 counts of VM.
Thus, as a whole, it is evident from the table that the presented work is much more suffi-
cient for load balancing during VM migration.

7 Discussions

The proposed DC- MFA model is easy to implement, it has few setting parameters, and
has fast convergence. The result section depicts the efficiency of the proposed model in
terms of sixfold multi objectives. All the objectives show less value than existing methods.
Sometimes, the existing methods possess suitable low values. But, in the end, the total cost
of the proposed model provides best values than the existing methods. Thus, our proposed
DC-MFA-based VM allocation model is suitable for cloud environments.

8 Conclusion

This paper had developed a novel approach for efficient load balancing approach by intro-
ducing DC-MFA, which was the extended version of the standard MFO algorithm. Here,
the VM selection under off-line consists of selecting one or more potential VMs for migra-
tion to reduce the resource load of the considered servers. This was achieved with DC-
MFA on the basis of sixfold objectives like Migration Cost, CPU utilization, security;
make span, migration cost, and resource cost. The performance of the proposed model was
analyzed by determining migration cost, energy consumption, response time, and secu-
rity analysis as well. On observing the total cost for 20 counts of VM allocated by a PM,
the best value of the presented work is 3.23%, 8.72%, 10.09%, 27.82%, 6.62%, and 9.61%
better than the traditional works like ABC + BA, ACO, CSA, KH, WOGA, and ILWOA
respectively for 20 counts of nodes.

Author’s Contributions All the authors have accepted responsibility for the entire content of this submitted
manuscript and approved submission.

Funding None.

Data Availability The data that support the findings of this study are openly available at https://​www.​kaggle.​
com/​discd​iver/​clouds.

Declarations
Conflict of interest The authors declare that they have no conflict of interest.

13
100 G. Verma

References
1. Annadanam, C. S., Chapram, S., & Ramesh, T. (2020). Intermediate node selection for Scatter-
Gather VM migration in cloud data center. Engineering Science and Technology, an International
Journal, in communication, 23(5), 989–997.
2. Patel, Y. S., Page, A., Nagdev, M., Choubey, A., & Das, S. K. (2020). On demand clock synchroni-
zation for live VM migration in distributed cloud data centers. Journal of Parallel and Distributed
Computing, 138, 15–31.
3. Sivagami, V. M., & Easwarakumar, K. S. (2019). An Improved Dynamic Fault Tolerant Manage-
ment Algorithm during VM migration in Cloud Data Center. Future Generation Computer Systems,
98, 35–43.
4. Mao, Bo., Yang, Y., Suzhen, Wu., Jiang, H., & Li, K.-C. (2019). IOFollow: Improving the perfor-
mance of VM live storage migration with IO following in the cloud. Future Generation Computer
Systems, 91, 167–176.
5. Caronia, F. P., Fiorelli, A., Zanchini, F., Santini, M., Monte, A. I. L., & Castorina, S. (2016).
Reconstruction with a pectoralis major myocutaneous flap after left first rib and clavicular chest
wall resection for a metastasis from laryngeal cancer. General thoracic and cardiovascular surgery,
64(5), 294–297.
6. Parisi, Giuseppe Fabio, Silvia Cutello, Giovanna Di Dio, Novella Rotolo, Mario La Rosa, and Sal-
vatore Leonardi. "Phenotypic expression of the p. Leu1077Pro CFTR mutation in Sicilian cystic
fibrosis patients." BMC research notes 6, no. 1 (2013): 1–5.
7. Fusini, F., Langella, F., Catani, O., Sergio, F., & Zanchini, F. (2017). Mini-invasive treatment for
brachymetatarsia of the fourth ray in females: Percutaneous osteotomy with mini-burr and external
fixation—a case series. The Journal of Foot and Ankle Surgery, 56(2), 390–394.
8. Sharma, Y., Si, W., Sun, D., & Javadi, B. (May 2019). Failure-aware energy-efficient VM consoli-
dation in cloud computing systems. Future Generation Computer Systems, 94, 620–633.
9. Aditya Bhardwaj, C. (2019). Rama krishna, “impact of factors affecting pre-copy vm migration
technique for cloud computing.” Materials Today: Proceedings, 18, 1138–1145.
10. Moghaddam, S. M., O’Sullivan, M., Walker, C., Piraghaj, S. F., & Unsworth, C. P. (2020). Embed-
ding individualized machine learning prediction models for energy efficient VM consolidation
within Cloud data centers. Future Generat Computer Sys, 106, 221–233.
11. Shirvani, M. H., Rahmani, A. M., & Sahafi, A. (2020). A survey study on VM migration and server
consolidation techniques in DVFS-enabled cloud datacenter: Taxonomy and challenges. Journal of
King Saud University - Computer and Information Sciences, in communication, 32(3), 2672–2686.
12. He, T. Z., Toosi, A. N., & Buyya, R. (2019). Performance evaluation of live VM migration in SDN-
enabled cloud data centers. Journal of Parallel and Distributed Computing, 131, 55–68.
13. Sudarshan Chakravarthy, A., Sudhakar, Ch., & Ramesh, T. (2019). Energy efficient VM scheduling
and routing in multi-tenant cloud data center. Sustainable Computing: Informatics and Systems, 22,
139–151.
14. Paulraj, G. J. L., Francis, S. A. J., Peter, J. D., & Jebadur, I. J. (2019). A combined forecast-based
VM migration in cloud data centers. Computers & Electrical Engineering, 69, 287–300.
15. Filho, M. C. S., Monteiro, C. C., Inácio, P. R. M., & Freire, M. M. (2018). Approaches for optimiz-
ing VM placement and migration in cloud environments: a survey. Journal of Parallel and Distrib-
uted Computing, 111, 222–250.
16. Wang, Z., Sun, D., Xue, G., Qian, S., & Li, M. (2019). Ada-Things: An adaptive VM monitoring
and migration strategy for internet of things applications. Journal of Parallel and Distributed Com-
puting, 132, 164–176.
17. Chen, Y.-R., & Li, J.-S. (2017). Staggered approach for alleviating TCP Incast in simultaneous
Multi-VM migration. Computer Communication, 106, 24–32.
18. Ray, S., & De Sarkar, A. (2012). Execution analysis of load balancing algorithms in cloud comput-
ing environment. International Journal on Cloud Computing: Services and Architecture IJCCSA,
2(5), 1–13.
19. Soni, A. (2015). Gagan Vishwakarma, and Yogendra Kumar Jain, “A bee colony based multi-objec-
tive load balancing technique for cloud computing environment.” International Journal of Comput-
ers and Applications, 114(4), 19–25.
20. Luo, J., Rao, L., & Liu, X. (2013). Temporal load balancing with service delay guarantees for data
center energy cost optimization. IEEE Transactions on Parallel and Distributed Systems, 25(3),
775–784.
21. George, A., & Rajakumar, B. R. (2013). Fuzzy aided ant colony optimization algorithm to solve
optimization problem. Intelligent Informatics. https://​doi.​org/​10.​1007/​978-3-​642-​32063-7_​23

13
Secure VM Migration in Cloud: Multi‑Criteria Perspective with… 101

22. Mirjalili, S. (2015). Moth-flame optimization algorithm: A novel nature-inspired heuristic para-
digm. Knowledge-Based Systems, 89, 228–249.
23. Rajakumar, B. R. (2018). Optimization using lion algorithm: A biological inspiration from
lion’s social behavior. Evolutionary Intelligence, Special Issue on Nature inspired algorithms
for high performance computing in computer vision, 11(1–2), 31–52. https://​doi.​org/​10.​1007/​
s12065-​018-​0168-y
24. Ninu Preetha, N. S., Brammya, G., Ramya, R., Praveena, S., Binu, D., & Rajakumar, B. R. (2018).
Grey wolf optimisation-based feature selection and classification for facial emotion recognition.
IET Biometrics, 7(5), 490–499. https://​doi.​org/​10.​1049/​iet-​bmt.​2017.​0160
25. Marsaline Beno, M., Valarmathi, I. R., Swamy, S. M., & Rajakumar, B. R. (2014). Threshold pre-
diction for segmenting tumour from brain MRI scans. International Journal of Imaging Systems
and Technology. https://​doi.​org/​10.​1002/​ima.​22087
26. Rajakumar, R. (2013). Impact of static and adaptive mutation techniques on genetic algorithm.
International Journal of Hybrid Intelligent Systems. https://​doi.​org/​10.​3233/​HIS-​120161
27. Z Guo, W Yao, D Wang 2017 A VM Migration Algorithm Based on Group Selection in Cloud Data
Center IFIP International Federation for Information Processing. Springer, cham
28. Karthikeyan, K., Sunder, R., Shankar, K., Lakshmanaprabu, S. K., Vijayakumar, V., Elhoseny, M.,
& Manogaran, G. (2018). Energy consumption analysis of VM migration in cloud using hybrid
swarm optimization (ABC–BA). The Journal of Supercomputing, 76(5), 3374–3390.
29. Sutar, S. G., Mali, P. J., & More, A. Y. (2020). Resource utilization enhancement through live VM
migration in cloud using ant colony optimization algorithm. International Journal of Speech Tech-
nology, 23, 79–85.
30. Liu, Y., Wang, K., Ge, L., Ye, L., & Cheng, J. (2019). Adaptive evaluation of vm placement and
migration scheduling algorithms using stochastic petri nets. IEEE Access, 7, 79810–79824.
31. Narayanan G.G., Saravanaguru, R.K., (2018). "Securing VM Migration Through IPSec Tunneling
and Onion Routing Algorithm," International Conference on Intelligent Computing and Control
Systems (ICICCS), Madurai, India, pp. 364–370,2018.
32. Torquato, M., Maciel, P., & Vieira, M. (2019). A Model for Availability and Security Risk Evalua-
tion for Systems With VMM Rejuvenation Enabled by VM Migration Scheduling. IEEE Access, 7,
138315–138326.
33. E. M. Kandoussi, I. El Mir, M. Hanini and A. Haqiq, "Modeling VM Migration as a Security Mech-
anism by using Continuous-Time Markov Chain Model," World Conference on Complex Systems
(WCCS), Ouarzazate, Morocco, pp. 1–6,2019.
34. Anurag Satpathy, Sourav Kanti Addya, Ashok Kumar Turuk, Banshidhar Majhi, Gadadhar Sahoo,
"Crow search based virtual machine placement strategy in cloud data centers with live migration",
Computers and Electrical Engineering, 2017
35. Minoo Soltanshahi, Reza Asemi, Nazi Shafiei, "Energy-aware virtual machines allocation by krill
herd algorithm in cloud data centers", Heliyon, vol.5, 2016.
36. Saxena, D., Gupta, I., Kumar, J., Singh, A. K., & Wen, X. (2021). A secure and multiobjective vir-
tual machine placement framework for cloud data center. IEEE Systems Journal. https://​doi.​org/​10.​
1109/​JSYST.​2021.​30925​21
37. Abdel-Basset, M., Abdle-Fatah, L., & Sangaiah, A. K. (2019). An improved Lévy based whale
optimization algorithm for bandwidth-efficient virtual machine placement in cloud computing envi-
ronment. Cluster Computing, 22(4), 8319–8334.
38. Vincenzo De Maio, Gabor Kecskemeti, Radu Prodan (2015). "A Workload-Aware Energy Model
for Virtual Machine Migration", IEEE International Conference on Cluster Computing, https://​doi.​
org/​10.​1109/​CLUST​ER.​2015.​47
39. S Chinprasertsuk, S Gertphol 2014 "Power Model for Virtual Machine in Cloud Computing",11th
International Joint Conference on Computer Science and Software Engineering, https://​doi.​org/​10.​
1109/​JCSSE.​2014.​68418​57
40. Vhat kar Kapil Netaji, Bhole GP 2020 "Optimal Container Resource Allocation Using Hybrid SA-
MFO Algorithm in Cloud Architecture", Multimedia Research, 3(1):11–20
41. Poluru, R. K., & Lokesh Kumar, R. (2019). Enhancement of ATC by Optimizing TCSC Configura-
tion using Adaptive Moth Flame Optimization Algorithm. Journal of Computational Mechanics,
Power System and Control, 2(3), 1–9.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

13
102 G. Verma

Garima Verma has done doctorate in Computer Science Engineering


from DIT University in 2018. She is UGC Net Qualified in Computer
Science, Gold Medalist in MTech, Computer Science and Engineering
from Uttarakhand Technical University. She has over 15 yrs of indus-
try and academic experience. She has published 18 Research papers in
various reputed peer reviewed journals and presented paper in various
conferences of international repute.

13

You might also like