Sie sind auf Seite 1von 20

Novel probabilistic resource migration

algorithm for cross-cloud live migration of


virtual machines in public cloud

Souvik Pal, Raghvendra Kumar, Le


Hoang Son, Krishnan Saravanan,
Mohamed Abdel-Basset, Gunasekaran
Manogaran & Pham Huy Thong
The Journal of Supercomputing
An International Journal of High-
Performance Computer Design,
Analysis, and Use

ISSN 0920-8542

J Supercomput
DOI 10.1007/s11227-019-02874-x

1 23
Your article is protected by copyright and
all rights are held exclusively by Springer
Science+Business Media, LLC, part of
Springer Nature. This e-offprint is for personal
use only and shall not be self-archived in
electronic repositories. If you wish to self-
archive your article, please use the accepted
manuscript version for posting on your own
website. You may further deposit the accepted
manuscript version in any repository,
provided it is only made publicly available 12
months after official publication or later and
provided acknowledgement is given to the
original source of publication and a link is
inserted to the published article on Springer's
website. The link must be accompanied by
the following text: "The final publication is
available at link.springer.com”.

1 23
Author's personal copy
The Journal of Supercomputing
https://doi.org/10.1007/s11227-019-02874-x

Novel probabilistic resource migration algorithm


for cross‑cloud live migration of virtual machines in public
cloud

Souvik Pal1 · Raghvendra Kumar2 · Le Hoang Son3,4 · Krishnan Saravanan5 ·


Mohamed Abdel‑Basset6   · Gunasekaran Manogaran7 · Pham Huy Thong8,9

© Springer Science+Business Media, LLC, part of Springer Nature 2019

Abstract
In cloud computing environment, cross-cloud live migration of virtual machines
(VMs) is a major concern in these days. Cloud computing provides the users with
huge, versatile and on-demand access to a bulk of customizable and configurable
registered physical devices or things. It helps organizations or enterprises to share
data efficiently by privately owned cloud or by the third-party servers. This type
of sharing of bulky data through cloud is more efficient and reliable. In an enter-
prise environment, one of the essential capabilities of cloud infrastructure is VM
migration. VM live migration basically involves the transference of instances that
includes the operating system, runtime memory pages and active CPU states from
source hub to the destination hub. In this paper, we have discussed on resource allo-
cation algorithm which performs better in utilization of CPU, time and memory. Our
proposed algorithm deals with the effective utilization of unoccupied memory, and
we have also measured VM memory stack flow of total memory for cloud comput-
ing architecture.

Keywords  Virtual machines · Virtualization · Virtual machines instance · Cross-


cloud

1 Introduction

Cloud computing is prominently and broadly utilized for empowering advanta-


geous, on-request arrange access to systems, servers, mass stockpiling and appli-
cations with negligible push to both specialist coop and end client under dis-
tributed computing [10]. For straightforwardness, a cloud is a framework that
comprises of an arrangement of physical processing assets, i.e., an arrangement

* Mohamed Abdel‑Basset
analyst_mohamed@yahoo.com
Extended author information available on the last page of the article

13
Vol.:(0123456789)
Author's personal copy
S. Pal et al.

of equipment, processors, memory, stockpiling, systems and data transmission,


which can be sorted into administrations [1]. With the emergence of web and
applications, virtualization gives a stage to upgrade tedious IT assets in a ver-
satile way [13]. Migration is the process of transferring a virtual machine (VM)
from one storage location to the other or to the host server [39]. In migration,
components like CPU, memory, networking and storage are all virtualized. Dur-
ing the migration of VMs, a set of simple portable data files are used for captur-
ing all states of VMs. In this regard, cross-cloud is a cloud that deals with several
hosts within a big cloud connected through LAN while transferring data [41].
From the user’s point of view, migration may lead to disconnection of getting
services from the cloud service provider (CSP). This is because of the lack of
active network connections or delay in workflow that make long time waiting,
while migration occurs [61]. To restrict this type of situation, a concept of live
migration which is known as the continuity and hot relocation in cloud comput-
ing condition was introduced [31]. When the VM is running in the source center
point or on the host server and amid live relocation, it is moved to the destina-
tion center point or other host server without mixing with any dynamic frame-
work associations. Live relocation is a streamlined way of making profitable use
of accessing CPU assets [40]. VM migration means transferring a VM starting
with one host physical machine to host machine that is either running or to be
booted up after placing the new VMs. Migration is required because of four rea-
sons: (1) lack of resources in the source center point; (2) when remote center is
running VMs on behalf of local center point because of dynamic unavailability
of resources; (3) for constraining the quantity of host machines that are running
remote destinations (less vitality utilization); and (4) for maximizing allocation of
VMs among the local machines rather than remote machines.
In order to handle the cross-cloud live migration of VMs in cloud computing,
many researches have been proposed in the literature as follows.
Phan et  al. [33] introduced different scheduling procedures for huge scale
constant information on cloud stages. Liu et  al. [26] highlighted the schedul-
ing issue for work process applications with security limitation. Jung et al. [20]
depicted a cloud bursting in view of maximally covered load-balancing algorithm
to streamline execution of huge information examination. Babukarthik et  al. [6]
proposed a half breed algorithm which centers on the diminishment of vitality
utilization. Tsai et al. [60] proposed a differential evolution algorithm. Chen et al.
[9] proposed a resource virtualization model for the most part giving computing
administrations. Sampaio and Barbosa [36] proposed a combination procedure to
enhance execution of dependability scheduling algorithms. Lin et  al. [25] pro-
posed a data transfer algorithm for distinct assignment scheduling in cloud com-
puting. Singh et al. [44] performed relative investigation of diverse algorithms to
cloud situation and proposed the half-and-half approach. Alkhanak et al. [4] con-
ducted a broad audit to research and dissected fundamental concepts of relevant
methodologies. Panda and Jana [32] introduced three task scheduling methods,
called MCC, MEMAX and CMMN for heterogeneous multi-cloud environment.
Stavrinides and Karatza [55] proposed a rundown scheduling heuristic in a het-
erogeneous PaaS cloud.

13
Author's personal copy
Novel probabilistic resource migration algorithm for…

Adhikary et al. [2] planned the task scheduling issue as a mixed integer linear
program (MILP) optimization. Zuo et  al. [68] proposed a task-situated multi-
objective scheduling method based on ant colony optimization to upgrade the
limited pool of open and private computing assets in a half-and-half cloud
computing scenario. Zuo et  al. [69] suggested a self-adaptive threshold-based
dynamically weighted load evaluation. Xiong et  al. [65] presented layered vir-
tual machine migration that divides cloud data center into different numbers of
regions according to bandwidth and utilization rate of hosts. Sofia and Ganesh
Kumar [45] proposed to control the vitality consumption viably. Xavier and
Annadurai [64] proposed chaotic social spider algorithm roused by social spi-
der to handle the issue of task scheduling in different heterogeneous virtual
machines. Mishra et  al. [30] suggested a method to map from VM to physical
machine according to resource requirement. Malekloo et al. [28] proposed VM
placement and VM consolidation. Callau-Zori et al. [8] argued that one critical
parameter is the host configuration by the number of VMs. Zhang et  al. [67]
have discussed regarding virtual machine migration linkage with the user mobil-
ity and optimization procedures. Wang and Gelenbe [63] have conferred the task
allocation schemes which meet the quality of service and have made a quantita-
tive comparison with round-robin scheduling. Fukai et  al. [12] have discussed
about BLMVisor which allows live migration on physical hardware components
and also evaluated the performance bare metal machine. Almutairi et al. [5] have
worked on data center sensitivity and proposed a virtual resource assignment
algorithm which support risk management. Hirofuchi et al. [19] have discussed
on live migration model on SimGrid VM which implements precopy algorithm.
They have also measured migration time, traffic and contention.
In this paper, we proposed a new approach to cross-cloud live migration of
VMs in cloud computing. As mentioned before, there are several techniques
to ensure live migration of virtual machines in cloud. However, none of the
research works presented in the past deals with the idea of probabilistic resource
mitigation. The need to introduce this resource allocation algorithm lies in the
fact that with cloud computing dealing with a lot of data, few of the attributes
that must be taken care of are better utilization of CPU, time and memory. Fur-
ther effective utilization of unoccupied memory as well as total memory must
be taken into consideration. Thus, the novelty lies in introduction of a proba-
bilistic resource migration algorithm that would ensure cross-cloud live migra-
tion of virtual machines in public cloud. To implement the same, we first esti-
mate unoccupied memory for all VMs and then evaluate the probability of each
VM having unoccupied memory. From the data that we obtain, we calculate the
probability for the required memory of incoming processes and find out pro-
cesses which have minimum difference from the value. If the difference is small
enough, it may be considered for migration. Similar processes are performed for
incoming processes.
The remainder of this paper is organized as follows. Section  2 introduces a
new algorithm for cross-cloud live migration. Section 3 discusses the result and
analysis. Section 4 highlights future researches and conclusions of this paper.

13
Author's personal copy
S. Pal et al.

2 Proposed Probabilistic Resource Migration Algorithm

In this section, we proposed a probabilistic resource migration algorithm to


achieve maximum utilization of unoccupied memory of VM. We describe the
general idea of the algorithm as well as its evaluation.

2.1 Probabilistic Resource Migration Algorithm

The idea is to migrate high unoccupied memory VMs from low one by proba-
bilistic. As mentioned, virtualization is the backbone of VM migration as it can
reduce energy consumption and integration of hardware cost. Migration of a VM
needs coordination in the transfer of source (S) and destination (D). It includes
the following three stages:
Stage 1 Controller and intermediate center points send migration demand to
destinations. D-center points can be in remote locales, and the demand can be
forwarded to another remote destination depending on resource availability.
Stage 2 The beneficiary may accept or reject upon the account capabilities and
usage of accounts. If all the business standards and accounting arrangement allow
the demand, an initiation transfer answer message will be sent to the asking for
center points serious I-centers.
Stage 3 The transfer operation from the S-center point to the D-center point
via I-centers and VMs on the source site will be migrated to destination locales.
Transfer initiation is done by the C-center point to D-center point and the transfer
is checked through token by the controller with the assistance of S-center point.
Figure 1 illustrates a scenario of migration.

Fig. 1  Scenario of migration

13
Author's personal copy
Novel probabilistic resource migration algorithm for…

The proposed probabilistic resource migration algorithm (PRMA) uses a prob-


ability of migration for each VM that have been estimated of unoccupied memory.
Besides, we estimate incoming resource requested by processes. Parameters are
occupied and unoccupied memory of VMs, probability of incoming memory request
and occupied memory by existing process of total memory of all VMs. The detail
description of the algorithm is mentioned below.

Algorithm 
Probabilistic resource migration algorithm (PRMA)
Input Process with the need of expected execution memory ( RMp)
Output Allocation of expected execution memory by VM migration

{ } [ ]
1. Estimate unoccupied memory for all VMs ∈ Vm1 , … , Vmn denoted as Munoc i ,
where i = 1, 2, 3, … , n ( i indicates the number of virtual machines.
[ ] [ [ ]]
Munoc i = TMi − MEP foreach Vm i , (1)
∑j � � � �
where MEP = 1 Mep j and Mep j is the occupied memory by each existing
process in VMs.
[ ]
Munoc i = TMi − Moc . (2)
2. Estimate the total unoccupied memory by taking integrated sum of estimated
unoccupied memory for each VM.
n
∑ [ ]
TMunoc = Munoc i . (3)
i=1

3. Estimate the probability of each VM having unoccupied memory.


[ ]
Munoc i
p(Vm)i = (4)
TMunoc

4. If the required memory for incoming process is greater than a threshold-TMunoc ,


we request rejected due to insufficient memory conditions and stop the algo-
rithm; otherwise, accept for memory allocation.
5. Calculate the probability for the required memory of incoming process.
RMp
p� (process) = . (5)
TMunoc
( )
6. Find out p Vm i which has minimum difference from the value of p� (process)
by
| � ( )|
|p (process) − p Vm i |.
| | (6)

13
Author's personal copy
S. Pal et al.

( ) | ( )|
7. Vm i is considered for migration if the difference of |p� (process) − p Vm i | is
| |
minimum. (( ) )
8. Check whether Munoc Vm i ≥ RMp . Continue particular incoming process
execution.
(( ) )
9. If Munoc (Vm )i < RMp , calculate the probability for migrating each existing
process in Vm i into another VM using
( )
PEP = Mep j ∕RMp . (7)

10. Find out process having less probability and consider it for migration into
another VM.
| |
11. Calculate the value of |Munoc − M(Process)migrate | for all VMs and consider the
| |
VM for accepting migration which gives minimum value.
12. Similar migration for existing process in VMs followed by steps 9 to 11.
13. Similar migration for new incoming process by steps 1 to 12.

2.2 Estimation evaluated by PRMA

The following steps give us an idea as to how the proposed algorithm may be
evaluated in order to carry live migration of virtual machines in public cloud. The
steps are detailed as follows.
[ ] [ ( ) [ ]]
Step 1 Munoc i = TMi − Memory occupied by existing process MEP in Vm i ..
Step 2 Estimate total unoccupied memory in all VMs.
Step 3 Estimate the probability of each VM having unoccupied memory. Now,
we consider the incoming process which is required resource memory 15.
Steps 4 and 5 Here, the required memory of the incoming process
( (RMp ) = 15 > TMunoc
This implies the process accepted to enter into VMs.
Step 6 Calculate the probability for required memory of the incoming process.
RM
p� (process) = TM p .
( )
unoc
Step 7 Find out p Vm i which is having minimum difference from the value of
( ).
p� (process)
Step 8 Vm 2 is considered
( )| for migration because we get the minimum value for
| �
|p (process) − p Vm 2 | in step 7.
| |
Steps 9 and 10(We)have to calculate the probability of each process occupied
the memory in Vm 2. ( ) ( )
Step 11 The process having Mep 2 is con to migrate into another VM if PEP 2
is smallest. ( )
Step 12 Consider Mep 2 will migrate to another VM. Now, we calculate( value )
| |
of |Munoc − M(Process)migrate | for all VMs. The smallest value for Mep 2
| |
ensures that the process in the corresponding VM is migrated.

13
Author's personal copy
Novel probabilistic resource migration algorithm for…

3 Experiments, results and discussions

In this section, we examine the experimental datasets that have been used for the
study. Based on the initial status of the VMs before migration, we perform the
experimental analysis and evaluate the proposed algorithm. Once we obtain the
results, we may conclude how the algorithm is effective in migrating live virtual
machines in public cloud. In order to strengthen our approach, we also perform a
comparative analysis of the existing works related to live migration of machines.

3.1 Experimental datasets

In this experiment, we consider the sample dataset in the form of VM memory con-
figuration, when incoming process demands memory for completing its execution.
It changes dynamically by incoming process requests (Table 1). We choose server
machine with the configuration of Intel Xeon CPU E5-2630 version 3, 32-Core Pro-
cessor, 128 GB DDR4 RAM. It can host almost 50 VMs in parallel. Memory unit
is measured in megabyte (MB) for this simulation environment. It is assumed that
all the VM are running in the same IP subnet. The simulation environment we are
working on is the CloudSim and Java Simulation scenarios. Total memory of each
VM, total occupied memory by existing process, unoccupied memory and memory
occupancy by each process are some parameters before migration. Table 1 shows the
initial values for all these parameters before migration.

3.2 Experimental analysis

Table 2 shows the incoming process request accepted by VM2 but insufficient mem-
ory. Here, in VM1 the total memory is of 30 and the total occupied memory by
existing process is 27. Therefore, unoccupied memory remaining is 3. The memory
occupancy by each process will be at {15,8,4}. In case of VM2 when total memory
of is 30 and the total occupied memory by existing process is 17, the unoccupied
memory remaining is 13 (insufficient memory condition). The memory occupancy
by each process will be at {15(incoming memory request),8,4}. In case of VM3, the
total memory is 25 and the total occupied memory by existing process is 17. The

Table 1  Initial status of VMs before migration


VM Memory
Total memory of Total occupied memory
( ) by Unoccupied
( ) Memory occupancy
each VM (TM)i existing process Moc i memory Munoc i (by each
) process
Mep i

VM1 30 27 3 {15, 8,4}


VM2 30 17 13 {10,7}
VM3 25 17 8 {12,5}

13
Author's personal copy
S. Pal et al.

Table 2  Incoming process request handled by VM2


VM Memory
Total memory of each Total occupied Unoccupied Memory occupancy
( )
VM (TM)i memory( by existing
) (memory
) by each process Mep i
process Moc i Munoc i

VM1 30 27 3 {15,8,4}
VM2 30 17 13 (Insufficient 15 (Incoming memory
Memory) request) {10,7}
VM3 25 17 8 {12,5}

unoccupied memory remaining is 8, and the memory occupancy by each process


will be at {12,5}.
The evaluation metrics recorded the probability of migration of each VM. Ini-
tially, we consider the memory status of three VMs that already have existing pro-
cess. We also consider another incoming process. Here, PRMA is used to migrate
the VM due to the need of execution memory of incoming process that may be
assigned or migrated upon the total memory allocation of virtual VMs. How the
unoccupied memory is assigned to VM by using PRMA algorithm is detailed as
follows:

3.3 Estimation evaluation by PRMA based on the experimental analysis

Based on the data we obtained from Tables 1 and 2, we can estimate the evaluation
by PRMA. [ We consider
] the[evaluation algorithm proposed in Sect. 2
( for) the
[ same.
]]
Step 1 Munoc i = TMi − Memory occupied by existing process MEP in Vm i
Thus, we can easily find the unoccupied memory in each VM.
[ ] [ ]
Munoc 1 = |30 − (15 + 8 + 4)| = 3; Munoc 2 = |30 − (10 + 7)| = 13
[ ] (8)
Vm3 = Munoc 3 = |25 − (12 + 5)| = 8.

Step 2 Estimate total unoccupied memory in all VMs.


{[ ] [ ] [ ]}
TMunoc = Munoc 1 + Munoc 2 + Munoc 3 = {3 + 13 + 8} = 24. (9)
Step 3 Estimate the probability of each VM having unoccupied memory.
[ ]
( ) Munoc 1 3 ( ) ( )
p Vm 1 = = = 0.125; similarly p Vm 2 = 0.541; p Vm 3 = 0.333.
TMunoc 24
(10)
Now, we consider the incoming process which is required resource memory 15.

13
Author's personal copy
Novel probabilistic resource migration algorithm for…

Steps 4 and 5 Here, the required memory of the incoming process


(RMp ) = 15 > TMunoc = 24.
This implies the process accepted to enter into VMs.
Step 6 Calculate the probability for required memory of the incoming process.
RMp 15
p� (process) = = = 0.625. (11)
TMunoc 24
( )
Step 7 Find out p Vm i which having minimum difference from the value of
p� (process).

| � ( ) | ( )
|p (process) − p Vm 1 | = 0.50; |p� (process) − p Vm 2
| |
| ( ) | (12)
= 0.084; |p� (process) − p Vm 3 | = 0.292.
| |
( )
Step 8 Vm 2 is considered for migration because we get the minimum value for
| � ( )|
|p (process) − p Vm 2 | in step 7.
| | (( ) ) (( ) )
Steps 9 and 10 Munoc Vm 2 = 13 and RMp = 15 , i.e., Munoc Vm 2 < RMp sat-
(isfied.
) We have to calculate
( ) the probability( )of each process
( occupied
) the memory in
Vm 2 . The process in Vm 2 occupied Mep 1 = 10 and Mep 2 = 7 , respectively.
( ) ( )
( ) Mep 1 10 ( ) Mep 2 7
PEP 1 = = = 0.666, Similarly PEP 2 = = = 0.466.
RMp 15 RMp 15
( ) (13)
Step 11
( The
) process having M ep 2
= 7 is considered to migrate into another VM
because PEP 2 is smallest. ( )
Step 12 We estimate the process having Mep 2 = 7 will migrate to another VM.
| |
Now, we calculate value of |Munoc − M(Process)migrate | for all VMs.
| |
| |
|Munoc − M(Process)migrate | = |3 − 7| = 4 for VM1,
| |
| |
|Munoc − M(Process)migrate | = |13 − 7| = 6 for VM2,
| | (14)
| |
|Munoc − M(Process)migrate | = |8 − 7| = 1 for VM3.
| |
( )
Hence, in this case the process in VM2 having Mep 2 = 7 is migrated to VM3.

3.4 Results and discussions

In this section, based on the data from Tables 1 and 2 and evaluation performed, we
discuss the results once the machines have been migrated. Table 3 shows the final
migrated status and process allocation in VMs. In case of VM1 when total mem-
ory of is 30 and the total occupied memory by existing process is 27, the unoc-
cupied memory remaining is 3. Then, the memory occupancy by each process will

13
Author's personal copy
S. Pal et al.

Table 3  Final migrated status and process allocation in VMs


VM Memory
Total memory of Total occupied memory
( ) by Unoccupied
( ) Memory occupancy
each VM (TM)i existing process Moc i memory Munoc i (by each
) process
Mep i

VM1 30 27 3 {15,8,4}
VM2 30 25 5 {15,10}
VM3 25 24 1 {7,12,5}

Table 4  Initial evaluation p(VM)s Probability value


metrics for each unoccupied of each VM
memory of VMs
p(VM)1 0.125
p(VM)2 0.541
p(VM)3 0.333

Table 5  Income process Income process Value of income process probability


probability evaluation metrics
p� (process) 0.625

Table 6  Evaluation metrics of Existing process in (VM)2 Probability of


(VM)2 each process
( )
PEP 0.666
( )1
PEP 0.466
2

be at {15,8,4}. In case of VM2 when total memory of is 30 and the total occupied
memory by existing process is 25, the unoccupied memory remaining is 5 (suffi-
cient memory condition). Then, the memory occupancy by each process will be at
{15,10}. In case of VM3 when total memory of is 25 and the total occupied memory
by existing process is 24, the unoccupied memory remaining is 1 then the memory
occupancy by each process will be at {7,12,5}.
Once we have the data from initial status and migrated status, we will perform
evaluation based on the same data. Tables 4 and 5 show the initial evaluation met-
rics. It is clear that (VM)2 will migrate, but it does not have sufficient memory to
accept incoming process. Thus, we have to calculate probability evaluation metrics
of existing process in (VM)2 . Table 6 shows the evaluation metrics of (VM)2 , and the
next process fully depended on this probability. Here, it is 0.466, i.e., the process
will migrate which occupied seven memories in (VM)2 to others VMs.
Figure  2 shows the maximum utilization of unoccupied memory. It is evident
that our proposed algorithm PRMA effectively migrates the existing unoccupied

13
Author's personal copy
Novel probabilistic resource migration algorithm for…

Fig. 2  Occupied memory of each VM before and after migration

memory. Probability ratio with the existing systems shows that PRMA has the high-
est probability in utilizing the unused/unoccupied memory in the VM.
Moreover, we also estimated the VM memory stack which maintains the uni-
form stack flow of the total memory for computing architecture. The implication of
PRMA reflects in Fig. 3 where VM memory stack view is described.
After estimation for first incoming process, the record will update. In this exper-
iment, the probability of acceptance for new incoming processes is mentioned in
Table  7. The probability of acceptance for new incoming process is high in VM2
(0.555) than other VMs. The high probability indicated VM2 is the next destined
VM for migration.

3.5 Comparative analysis

In past, several approaches have been adapted for live migration of machines
in cloud computing environments. Kapil et  al. [21] presented some live vir-
tual machine migration techniques. The metrics used were preparation time,

Fig. 3  Described stack view of unoccupied memory for each VM

13

13
Table 7  Process status in VMs
VM Before migration After migration
Process assigned Probability Process assigned New process Probability of acceptance
assigned for new incoming process

VM1 p1 , p2 , p3 Vm 1 => {15, 8, 4} 0.125 p1 , p2 , p3 Vm 1 => {15, 8, 4} No 0.333


(( ) ) (( ) )

VM2 p1 , p2 Vm 2 => {10, 7} 0.541 p1 , p2 Vm 2 => {15, 10} Yes 0.555


(( ) ) (( ) )

VM3 p1 , p2 Vm 3 => {12, 5} 0.333 p1 , p2 , p3 Vm 3 => {7, 12, 5} Yes 0.111


(( ) ) (( ) )
Author's personal copy
S. Pal et al.
Author's personal copy
Novel probabilistic resource migration algorithm for…

downtime, pages transferred, application degradation, etc., whereas the tech-


niques included postcopy approaches, precopy approaches and multiple machine
migrations. The limitations were lack of technical implementation as well as var-
ying types of bandwidth and workloads. Breitgand et al. [7] proposed cost-aware
live migration services in cloud using fixed bandwidth migration and variable
bandwidth migration techniques. The performance measure is insufficient due to
bandwidth issues and change in request rate. Malik and Barde [29] used predic-
tion of CPU usage as a method of live migration of virtual machines in cloud.
The number of migrations, percentage of service-level agreement (SLA) viola-
tions and energy consumption were analyzed for the same. Workload is fixed in
this case; otherwise, it leads to overload and SLA violations. Agarwal and Raina
[3] relied on the Red Hat Cluster Suite services for live migration of machines
in cloud. System architecture is proposed to carry out the same; however, the
study lacks technical implementation. Ye et  al. [66] used resource reservation
in cloud computing environment for live migration. A live migration framework
along with experimental analysis was exhibited to promote the efficiency of the
system. The workload performance, downtime and migration time were evalu-
ated. Performance overheads as well as migration failures exist for the conducted
research work. Our proposed work introduces a novel algorithm, i.e., probabilistic
resource migration algorithm to conduct live migration of virtual machines in the
cloud. Based on the experimental analysis, where we have performed evaluations
based on dataset and our proposed evaluation algorithm, it is evident that our pro-
posed algorithm PRMA effectively migrates the machines.
Figure  4 shows the comparative analysis between the existing and the pro-
posed works. The parameters considered here are CPU, memory utilization and
time against the utilization rate with time. We find that the proposed algorithms

Fig. 4  Comparative analysis with existing works

13
Author's personal copy
S. Pal et al.

perform better in the case utilization of CPU, memory and time when the number
of virtual machines increases.

4 Conclusions and future scope

With regard to distributed computing, virtualization utilizes PC assets to mimic the


entire computers. We have exhibited a way to deal with cross-cloud live migration
of VMs in cloud computing. Virtualization gives adaptable and dynamic asset pro-
visioning capacities to end clients. Live movement of virtual machines exchanges
the condition of VM starting with one host to the next without intruding on the run-
ning VMs. Relocation of VMs mitigates overburden state of server farms and keeps
up the continuous administrations. The primary goal of the relocation procedure is
to accomplish less downtime and movement time. The VM relocation incorporates
the transmission of CPU, memory and input/output to the objective machine. By
relocating the whole state and the memory substance of VMs to others, we encour-
age quick development and sharing of intelligent workloads. Negligible downtime
additionally makes quick live movement such that the client has not to sit tight for a
drawn out stretch of time in cloud computing environment.
In the future, virtual migration services take place in the research field to achieve
its objectives, and there are different potential areas for research as follows.

• One of the budding areas that have enough significance is live migration between
different subnet IPs. Our proposed algorithm works within only same IP subnet.
Such kind of future work may facilitate better live migration of VMs in the WAN
scenario.
• In live migration, security aspects are also considerable point in real-time appli-
cations. In live migration, transmission medium security is important. Hence, to
make a secure medium future work may be done by creating secure VPN, SSL
and IPSEC. A better algorithm can be used in the firewall to ensure network and
perimeter security.
• To make secure live migration, client-level security is also needed. Future
research works can be carried out by making input validation, SSH and open
API. This kind of work may protect from different attacks like broken authenti-
cation, SQL injection and intrusion.
• More advanced machine learning algorithms can be applied for VM in cloud
computing [11, 14–18, 22–24, 27, 34, 35, 37, 38, 42, 43, 46–54, 56–59, 62].

References
1. Abderrahim W, Choukair Z (2017) The three-dimensional model for dependability integration in
cloud computing. Ann Telecommun 72(5–6):371–384

13
Author's personal copy
Novel probabilistic resource migration algorithm for…

2. Adhikary T, Das AK, Razzaque MA, Almogren A, Alrubaian M, Hassan MM (2016) Quality of ser-
vice aware reliable task scheduling in vehicular cloud computing. Mob Netw Appl 21(3):482–493
3. Agarwal A, Raina S (2012) Live migration of virtual machines in cloud. Int J Sci Res Publ
2(6):45–52
4. Alkhanak EN, Lee SP, Khan SUR (2015) Cost-aware challenges for workflow scheduling
approaches in cloud computing environments: taxonomy and opportunities. Future Gener Comput
Syst 50:3–21
5. Almutairi A, Sarfraz MI, Ghafoor A (2018) Risk-aware management of virtual resources in access
controlled service-oriented cloud datacenters. IEEE Trans Cloud Comput 6(1):168–181
6. Babukarthik RG, Raju R, Dhavachelvan P (2012) Energy-aware scheduling using hybrid algorithm
for cloud computing. In: 2012 Third International Conference on Computing Communication &
Networking Technologies (ICCCNT). IEEE, pp 1–6
7. Breitgand D, Kutiel G, Raz D (2011) Cost-aware live migration of services in the cloud. In: Hot-
ICE’11 Proceedings of the 11th USENIX Conference on Hot Topics in Management of Internet,
cloud, and Enterprise Networks and Services, p 3
8. Callau-Zori M, Samoila L, Orgerie AC, Pierre G (2017) An experiment-driven energy consumption
model for virtual machine management systems. Sustain Comput Inform Syst 18:163–174
9. Chen X, Zhang J, Li J, Li X (2013) Resource virtualization methodology for on-demand allocation
in cloud computing systems. SOCA 7(2):77–100
10. Dave A, Patel B, Bhatt G (2016) Load balancing in cloud computing using optimization techniques:
a study. In: International Conference on Communication and Electronics Systems (ICCES). IEEE,
pp 1–6
11. Doss S, Nayyar A, Suseendran G, Tanwar S, Khanna A, Son LH, Thong PH (2018) APD-JFAD:
accurate prevention and detection of jelly fish attack in MANET. IEEE Access 6:56954–56965
12. Fukai T, Shinagawa T, Kato K (2018) Live migration in bare-metal clouds. IEEE Trans Cloud Com-
put. https​://doi.org/10.1109/TCC.2018.28489​81
13. Gai K, Qiu M, Zhao H (2016) Cost-aware multimedia data allocation for heterogeneous memory
using genetic algorithm in cloud computing. IEEE Trans Cloud Comput. https​://doi.org/10.1109/
TCC.2016.25941​72
14. Giap CN, Son LH, Chiclana F (2018) Dynamic structural neural network. J Intell Fuzzy Syst
34:2479–2490
15. Hai DT, Son H, Vinh LT (2017) Novel fuzzy clustering scheme for 3D wireless sensor networks.
Appl Soft Comput 54:141–149
16. Hemanth DJ, Anitha J, Son LH (2018) Brain signal based human emotion analysis by circular back
propagation and Deep Kohonen neural networks. Comput Electr Eng 68:170–180
17. Hemanth DJ, Anitha J, Son LH, Mittal M (2018) Diabetic retinopathy diagnosis from retinal images
using modified Hopfield neural network. J Med Syst 42(12):247
18. Hemanth J, Anitha J, Naaji A, Geman O, Popescu D, Son LH (2018) A Modified deep convolutional
neural network for abnormal brain image classification. IEEE Access 7(1):4275–4283
19. Hirofuchi T, Lebre A, Pouilloux L (2018) SimGrid VM: virtual machine support for a simulation
framework of distributed systems. IEEE Trans Cloud Comput 6(1):221–234
20. Jung G, Gnanasambandam N, Mukherjee T (2012) Synchronous parallel processing of big-data ana-
lytics services to optimize performance in federated clouds. In: 2012 IEEE 5th International Confer-
ence Cloud Computing (CLOUD). IEEE, pp 811–818
21. Kapil D, Pilli E, Joshi R (2013) Live virtual machine migration techniques: survey and research
challenges. In: 2013 3rd IEEE International Advance Computing Conference (IACC), p p 78–83
22. Kapoor R, Gupta R, Kumar R, Son LH, Jha S (2019) New scheme for underwater acoustically wire-
less transmission using direct sequence code division multiple access in MIMO systems. Wirel
Netw. https​://doi.org/10.1007/s1127​6-018-1750-z
23. Kapoor R, Gupta R, Son LH, Jha S, Kumar R (2018) Boosting performance of power quality event
identification with KL divergence measure and standard deviation. Measurement 126:134–142
24. Kapoor R, Gupta R, Son LH, Jha S, Kumar R (2018) Detection of power quality event using histo-
gram of oriented gradients and support vector machine. Measurement 120:52–75
25. Lin W, Liang C, Wang JZ, Buyya R (2014) Bandwidth-aware divisible task scheduling for cloud
computing. Softw Pract Exp 44(2):163–174
26. Liu H, Abraham A, Snášel V, McLoone S (2012) Swarm scheduling approaches for work-flow
applications with security constraints in distributed data-intensive computing environments. Inf Sci
192:228–243

13
Author's personal copy
S. Pal et al.

27. Long HV, Ali M, Khan M, Tu DN (2019) A novel approach for fuzzy clustering based on neutro-
sophic association matrix. Comput Ind Eng. https​://doi.org/10.1016/j.cie.2018.11.007
28. Malekloo MH, Kara N, El Barachi M (2018) An energy efficient and SLA compliant approach for
resource allocation and consolidation in cloud computing environments. Sustain Comput Inform
Syst 17:9–24
29. Malik V, Barde C (2015) Live migration of virtual machines in cloud environment using prediction
of CPU usage. Int J Comput Appl 117(23):124–131
30. Mishra SK, Puthal D, Sahoo B, Jayaraman PP, Jun S, Zomaya AY, Ranjan R (2018) Energy-efficient
VM-placement in cloud data center. Sustain Comput Inform Syst 20:48–55
31. Osanaiye O, Chen S, Yan Z, Lu R, Choo K, Dlodlo M (2017) From cloud to fog computing: a
review and a conceptual live VM migration framework. IEEE Access 5:8284–8300
32. Panda SK, Jana PK (2015) Efficient task scheduling algorithms for heterogeneous multi-cloud envi-
ronment. J Supercomput 71(4):1505–1533
33. Phan LT, Zhang Z, Zheng Q, Loo BT, Lee I (2011) An empirical analysis of scheduling techniques
for real-time cloud-based data processing. In: 2011 IEEE International Conference on Service-Ori-
ented Computing and Applications (SOCA). IEEE, pp 1–8
34. Phuong PTM, Thong PH, Son LH (2018) Theoretical analysis of picture fuzzy clustering: conver-
gence and property. J Comput Sci Cybern 34(1):17–32
35. Robinson YH, Julie EG, Saravanan K, Kumar R, Son LH (2019) FD-AOMDV: fault-tolerant dis-
joint ad-hoc on-demand multipath distance vector routing algorithm in mobile ad-hoc networks. J
Ambient Intell Humaniz Comput. https​://doi.org/10.1007/s1265​2-018-1126-3
36. Sampaio AM, Barbosa JG (2013) Optimizing energy-efficiency in high-available scientific cloud
environments. In: 2013 Third International Conference on Cloud and Green Computing (CGC).
IEEE, pp 76–83
37. Saravanan K, Anusuya E, Kumar R, Son LH (2018) Real-time water quality monitoring using Inter-
net of Things in SCADA. Environ Monit Assess 190(9):556
38. Saravanan K, Aswini S, Kumar R, Son LH (2019) How to prevent maritime border collision for fish-
eries? A design of real-time automatic identification system. Earth Sci Inf. https​://doi.org/10.1007/
s1214​5-018-0371-5
39. Satpathy A, Addya SK, Turuk AK, Majhi B, Sahoo G (2017) Crow search based virtual machine
placement strategy in cloud data centers with live migration. Comput Electr Eng. https​://doi.
org/10.1016/j.compe​lecen​g.2017.12.032
40. Seo D, Jeon YB, Lee SH, Lee KH (2016) Cloud computing for ubiquitous computing on M2 M and
IoT environment mobile application. Clust Comput 19(2):1001–1013
41. Sharma P, Lee S, Guo T, Irwin D, Shenoy P (2018) Managing risk in a derivative IaaS cloud. IEEE
Trans Parallel Distrib Syst 29(8):1750–1765
42. Singh K, Singh K, Son LH, Aziz A (2018) Congestion control in wireless sensor networks by hybrid
multi-objective optimization algorithm. Comput Netw 138:90–107
43. Singh N, Son LH, Chiclana F, Jean-Pierre M (2019) A new fusion of salp swarm with sine cosine
for optimization of non-linear functions. Engineering with Computers. https​://doi.org/10.1007/
s0036​6-018-00696​-8
44. Singh RM, Paul S, Kumar A (2014) Task scheduling in cloud computing. Int J Comput Sci Inf
Technol: IJCSIT 5(6):7940–7944
45. Sofia AS, Ganesh Kumar P (2018) Multi-objective task scheduling to minimize energy consumption
and makespan of cloud computing using NSGA-II. J Netw Syst Manage 26(2):463–485
46. Son LH (2015) A novel kernel fuzzy clustering algorithm for geo-demographic analysis. Inf Sci
Inform Comput Sci Intell Syst Appl Int J 317(C):202–223
47. Son LH (2016) Generalized picture distance measure and applications to picture fuzzy clustering.
Appl Soft Comput 46(C):284–295
48. Son LH, Hai PV (2016) A novel multiple fuzzy clustering method based on internal clustering vali-
dation measures with gradient descent. Int J Fuzzy Syst 18(5):894–903
49. Son LH, Jha S, Kumar R, Chatterjee JM, Khari M (2019) Collaborative handshaking approaches
between internet of computing and internet of things towards a smart world: a review from 2009–
2017. Telecommun Syst. https​://doi.org/10.1007/s1123​5-018-0481-x
50. Son LH, Tien ND (2017) Tune up fuzzy C-means for big data: some novel hybrid clustering algo-
rithms based on initial selection and incremental clustering. Int J Fuzzy Syst 19(5):1585–1602
51. Son LH, Tuan TM (2016) A cooperative semi-supervised fuzzy clustering framework for dental
X-ray image segmentation. Expert Syst Appl 46:380–393

13
Author's personal copy
Novel probabilistic resource migration algorithm for…

52. Son LH, Fujita H (2019) Neural-fuzzy with representative sets for prediction of student perfor-
mance. Appl Intell 49(1):172–187
53. Son LH, Thong PH (2017) Some novel hybrid forecast methods based on picture fuzzy clustering
for weather nowcasting from satellite image sequences. Appl Intell 46(1):1–15
54. Son LH, Tuan TM (2017) Dental segmentation from X-ray images using semi-supervised fuzzy
clustering with spatial constraints. Eng Appl Artif Intell 59:186–195
55. Stavrinides GL, Karatza HD (2015) A cost-effective and qos-aware approach to scheduling real-
time workflow applications in paas and saas clouds. In: 2015 3rd International Conference on Future
Internet of Things and Cloud (FiCloud). IEEE, pp 231–239
56. Tam NT, Hai DT, Son LH, Vinh LT (2018) Improving lifetime and network connections of 3D
wireless sensor networks based on fuzzy clustering and particle swarm optimization. Wireless Netw
24(5):1477–1490
57. Thong PH, Son LH (2016) Picture fuzzy clustering: a new computational intelligence method. Soft
Comput 20(9):3549–3562
58. Thong PH, Son LH (2016) A novel automatic picture fuzzy clustering method based on particle
swarm optimization and picture composite cardinality. Knowl-Based Syst 109:48–60
59. Thong PH, Son LH (2016) Picture fuzzy clustering for complex data. Eng Appl Artif Intell

56:121–130
60. Tsai JT, Fang JC, Chou JH (2013) Optimized task scheduling and resource allocation on cloud
computing environment using improved differential evolution algorithm. Comput Oper Res
40(12):3045–3055
61. Tsakalozos K, Verroios V, Roussopoulos M, Delis A (2017) Live VM migration under time-con-
strains in share-nothing IaaS-clouds. IEEE Trans Parallel Distrib Syst 28(8):2285–2298
62. Tuan TM, Ngan TT, Son LH (2016) A novel semi-supervised fuzzy clustering method based on
interactive fuzzy satisficing for dental X-ray image segmentation. Appl Intell 45(2):402–428
63. Wang L, Gelenbe E (2018) Adaptive dispatching of tasks in the cloud. IEEE Trans Cloud Comput
6(1):33–45
64. Xavier VA, Annadurai S (2018) Chaotic social spider algorithm for load balance aware task sched-
uling in cloud computing. Clust Comput. https​://doi.org/10.1007/s1058​6-018-1823-x
65. Xiong et al (2017) Layered virtual machine migration algorithm for network resource balancing in
cloud computing. Front Comput Sci 8(2):187–198
66. Ye K, Jiang X, Huang D, Chen J, Wang B (2011) Live migration of multiple virtual machines with
resource reservation in cloud computing environments. In: 2011 IEEE 4th International Conference
on Cloud Computing, pp 48–53
67. Zhang F, Liu G, Fu X, Yahyapour R (2018) A survey on virtual machine migration: challenges,
techniques, and open issues. IEEE Commun Surv Tutor 20(2):1206–1243
68. Zuo L, Shu L, Dong S, Chen Y, Yan L (2017) A multi-objective hybrid cloud resource scheduling
method based on deadline and cost constraints. IEEE Access 5:22067–22080
69. Zuo L, Shu L, Dong S, Zhu C, Zhou Z (2017) Dynamically weighted load evaluation method based
on self-adaptive threshold in cloud computing. Mob Netw Appl 22(1):4–18

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

Affiliations

Souvik Pal1 · Raghvendra Kumar2 · Le Hoang Son3,4 · Krishnan Saravanan5 ·


Mohamed Abdel‑Basset6   · Gunasekaran Manogaran7 · Pham Huy Thong8,9
Souvik Pal
souvikpal22@gmail.com
Raghvendra Kumar
raghvendraagrawal7@gmail.com

13
Author's personal copy
S. Pal et al.

Le Hoang Son
sonlh@vnu.edu.vn
Krishnan Saravanan
saravanan.krishnann@gmail.com
Gunasekaran Manogaran
gmanogaran@ucdavis.edu
Pham Huy Thong
phamhuythong@tdtu.edu.vn
1
Department of Computer Science and Engineering, JIS College, Kalyani, India
2
Department of Computer Science and Engineering, LNCT College, Jabalpur, MP, India
3
Institute of Research and Development, Duy Tan University, Da Nang 550000, Vietnam
4
VNU Information Technology Institute, Vietnam National University, Hanoi, Vietnam
5
Department of Computer Science and Engineering, Anna University Regional Campus,
Tirunelveli, India
6
Department of Operations Research, Faculty of Computers and Informatics, Zagazig University,
Zagazig, Sharqiyah, Egypt
7
University of California, Davis, USA
8
Division of Data Science, Ton Duc Thang University, Ho Chi Minh City 700000, Vietnam
9
Faculty of Information Technology, Ton Duc Thang University, Ho Chi Minh City 700000,
Vietnam

13

Das könnte Ihnen auch gefallen