IRJET- Research Paper on Energy-Aware Virtual Machine Migration for Cloud Computing

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 02 | Feb 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 114
Research Paper On ENERGY-AWARE VIRTUAL MACHINE MIGRATION
FOR CLOUD COMPUTING
Ms. Ramandeep kaur, Dr. Vijay Kumar Joshi2
1M.Tech Scholar
2Principal
---------------------------------------------------------------------***------------------------------------------------------------------
Abstract:- In cloud computing, users tries to migrate the computations to the disperse environment that can use number
of resources for completing execution rapidly. Virtualization can be defined on the basis of varied VMs (Virtual machines)
on one PM (Physical machine). In the procedure of migration, VM moves one PM to other. In case of offline migration,
procedure halts when VM continues on target machine, whereas, in case of live migration, procedure can be implemented
without any interruption. Live migration is the migration process and the virtual machine always seems to respond to it
from the perspective of the customer. In the data centre, real-time migration of virtual machines plays an important role.
Live migration is widely used for load balancing, energy reduction, and dynamic resizing to increase availability and
hardware maintenance. In this research, GA (Genetic algorithm) is used for optimization and ANN (Artificial neural
network) for classification. Numbers of QoS parameters are used to evaluate the performance.
Keywords: Cloud computing, VMM (Virtual machine migration), GA (Genetic algorithm), ANN (Artificial Neural Network),
SLA violation, Energy consumption
1. Introduction
With the advent of virtualization, the computation paradigm viz. Cloud computing has achieved huge success. The data
centres of cloud utilize virtualization and host of VMs (virtual machines) in PM (Physical machines) with effective resource
usage [1]. For the evolution of data centers, effective resource usage plays a significant role. With the data centres, one may
lessen the hardware as well the computational cost with the environmental and energy consumption issues. The main
objective of data centres is dependent on two types, namely, Migration and Virtualization.
VMs may shift from one PM to another. Generally, migration is an important tool for administrator of clusters and data
centres. VM migration is apparent to modern technology of virtualization and applications that supports it. Migration
occurs with the packaging cost of the data by means of VMs for migration and transferring that towards network from the
PM source towards Destination PM [2]. Other issue with VM migration is the required workload hotspot detection which
has to be migrates initially. Because of this issue, initialization of migration is considered as a manual task. Initiated
manually migration may lack in essential reactivity for responding to abrupt workload variations and was error prone, as
every re-shuffle may need migration or changing of number of VMs to re-balance the system load. For these issues, the
academic researchers has emphasised on the improvement and making the process automatic for VM migration in data
clusters and centres [3].
The issue of VM migration is certainly a decision making issue. On the basis of some response received from data centre,
the policy has to fix when, how and which VMs need to be migrated on the basis of varied aims [4]. According to the
solution of this problem and on the basis of realistic observations that the technology need to evaluate live migration, it is
probable to consider the queries of “When to migrate”, “Which VMs should be migrated”. It is definitely a control issue and
the technology provides live migration (actuator) that may be utilized for satisfying the objective of data centres owner.
Number of probable formulations are there of this issue which are dependent on modelling methods, like, equations,
queuing network, Markov models [5].
1.1 Virtualization
Virtualization is defined as the development of virtual resources like desktop, server, file, OS (Operating System), network
or storage. The main aim of virtualization is the management of workloads by completely changing the existing computing
for making it more scalable [6]. The general type of virtualization is OS level virtualization. In OS level virtualization, it is
mandatory to execute variety of OS on single hardware piece. Virtualization divides physical software and hardware by
following hardware with software. When varied OS operates on main OS by virtualization, then it is termed as VM (Virtual
machine). VM is nothing but a form of data file on physical machine which could be moved and reproduced to some other
computer like a normal data file. The machines in virtual environment utilize two types of file structures, namely,
hardware and hard drive. The virtualization software (hypervisor) provides caching technology that could be utilized to

cache the variations to the virtual hardware (virtual hard disk). This technology lets the user to abandon the changes being
executed to the OS and allows it to boot from the pre-defined state [7].
Figure 1: Virtualization
1.2 VM Migration
VM is referred as the instance of OS with more application executing in inaccessible partitioning in the machine. There
would be number of machines that may be running on the top of lone PM. When one physical host become overloaded, it
may be needed to dynamically send some amount of the load to other machine with less interruption to the user [8]. This
procedure of transferring a VM from one physical host to other is known as Migration. Previously, for transferring a VM in
two physical hosts, it was mandatory to close the VM and the allocation was required for sending the resources to novel
physical host. VM consists of two types of techniques [9]:
i. Live migration
It can be describes as the changing of VM from one physical host to another being powered on. When the process is
carrying out properly, this procedure can be executed without some conspicuous effect from the end user’s point of view.
ii. Regular migration
It is the migration of the powered off VM. With this, the user has the option to transfer the linked disks from one data store
to other. VMs are not required for being on shared storage.
Figure 1.1: VM Migration of two operating systems
In the past, many researchers have been working on formulating energy-saving algorithms that reduce energy
consumption. Many algorithms save data centre energy by shutting down or by placing idle servers in the server's sleep

mode. Researchers have proposed a minimal migration (MM) virtual machine placement algorithm that takes into account
the host CPU utilization of the list of virtual machines in descending order of CPU utilization [10].
The performance of the algorithms was better, but the SLA parameters are not considered when selecting a virtual
machine for migration, which may be affected by live migration [11]. Most violations occur during live migration of virtual
machines, migrating parameters (such as availability, response time, throughput, network bandwidth, etc.) that affect SLA
(Service level agreement). Therefore, a new method for developing SLA-aware energy-saving algorithms needs to be
allocated for data centre resources. The concept of virtual machine (VM) is related to the reduction of energy utilization
because it essentially reduces the idle power of the general base [12].
2. Proposed Architecture
In the past few years, the layout and migration of VMs has always been a tough task. Whenever a physical machine cannot
meet the full needs of a VM, there is a need for VMM. During this process, the VM will migrate without interfering with the
jobs in the running state. In sharp contrast to this, many researchers have made efforts to minimize SLA violations and the
number of migrations in various algorithms. The previously implemented algorithms are complex in nature and take a lot
of time to find and allocate physical machines, as SLAs are also violated. Therefore, to improve SLA violations and
minimize the number of migrations, ANN and GA has been used in this research to optimize VM migration.
2.1 ANN ( Artificial Neural Network)
Neural network is a computational system inspired by the structure, processing method and learning ability of the brain.
In this, there are very large numbers of neurons like processing elements and between those elements weights are there
which connects the processing elements. Knowledge is acquired by the network through a learning process. Artificial
neural network consists of simple computational units called neurons, which are connected to each other. They are highly
used because of its highly complex problem solving nature. A main feature of these networks is their adaptive nature,
where “learning by example” replaces “programming” in processing of problems. This feature makes it more appealing in
the application domains as person can have little or no knowledge about the problem to be solved but training data is
readily available. In the area of classification and prediction, ANN is highly used. ANN algorithm is defined below:
3. Artificial neural network algorithm
Assign ANN in the network
Describe input parameters viz. Neuron, training data with the VMs properties and group in the training data.
Develop ANN structure of ANN with initialization
Net = newff (Training data, Group, Neurons) Define the training metrics of ANN like number of iterations, parameters,
transfer function, O/P goal and training method.
Train the ANN with the parameters by Net = train (Net, Training data, Group) Organize the VM migrations with network
simulation Migrated_VM=sim(Net, Test VMs)
Develop the list of migrated VMs
3.1 GA (Genetic Algorithm)
GA (Genetic Algorithm) is basically used in the search space of large-scale applications. The advantage of the genetic
algorithm is that the process is fully automated and local minima are avoided. The main components of a genetic algorithm
are: crossover, mutation and fitness functions. Chromosome displays genetic algorithm solution. Crossover operations are
used to generate new chromosomes from the parent set, and mutation operators add mutations. The fitness function
performs chromosomes according to defined criteria. Improved chromosome fitness increases their chances of survival.
The population is a collection of chromosomes. GA algorithm is defined below:
Genetic Algorithm
Create an arbitrary population of n chromosomes
Compute the fitness of each chromosome in the population

Build a new population with a reiteration of underneath steps till the new population is complete [Selection] Selection of
two parent chromosome from the population according to fitness. [Crossover] Cross over the parents for the
development of new offspring with crossover probability
[Mutation] Mutate the new offspring at each locus with mutation probability [Accepting] Place new offspring in a new
population
Use newly generated population for the subsequent algorithm execution
If the end condition is fulfilled, halt and return the better solution in current population
As this research deals with the designing of an algorithm for analyzing the live migration behaviour in cloud data centres
and for that ANN is used as a classification algorithm and GA as an Optimization algorithm. Therefore, the workflow of the
work has been designed and is shown in below figure.
Figure 1.3: Proposed Workflow
4. Results and Discussion
In this section, the results obtained after the simulation of the work are defined. For the experimentation, parameters,
namely, Energy consumption, number of jobs completed and total number of migrations are calculated.
Table 1: Energy consumption evaluation
Number of Iterations Energy consumption (KWh)
1 0.0789
2 0.0438
3 0.0088
4 0.146
5 0.175

Figure 1.4: Energy consumption
Table 1 and figure 1.4 define the results obtained after the evaluation of simulation model. As depicted in figure 1.4, a
graph has been drawn that shows the utilization of energy consumption by means of number of iterations. X-axis in the
figure defines the number of iterations and Y-axis defines the values obtained after the evaluation. For an efficient system,
utilization of energy should be less and the energy consumption in the proposed work is 0.0905 approximately.
Table 1.1: Number of migrations evaluation
Number of Iterations Number of migrations
1 3
2 1
3 1
4 1
5 4
Figure 1.5: Number of migrations
Table 1.1 and figure 1.5 define the results obtained after the evaluation of simulation model. As depicted in figure 1.5, a
graph has been drawn that shows the number of migrations by means of number of iterations by artificial neural network
and genetic algorithm. X-axis in the figure defines the number of iterations and Y-axis defines the values obtained after the
evaluation. The average value obtained for number of migration in the proposed system is 2.

Table 1.2: SLA Violation evaluation
Number of Iterations SLA violation
1 0.286
2 0.624
3 1.681
4 0.881
5 0.959
Figure 1.6: SLA violation
Table 1.2 and figure 1.6 defines the results obtained after the evaluation of simulation model. As depicted in figure 1.6, a
graph has been drawn that shows the number of migrations by means of number of iterations by artificial neural network
and genetic algorithm. X-axis in the figure defines the number of iterations and Y-axis defines the values obtained after the
evaluation. The average value obtained for SLA violation in the proposed system is 0.8862.
5. Conclusion
VM migration and VM placement has always being a challenging task from a long. Number of researchers has executed
their work for minimizing SLA violation and number of migration with energy consumption in number of ways. The
existing algorithms are composite in nature and usually take a lot of time for allocating and finding a PM. For the effective
utilization, GA (Genetic algorithm) as optimization algorithm and ANN (Artificial neural network) as classification
algorithm has been used. The work has been executed by five number of iterations. For an efficient system, utilization of
energy should be less and the energy consumption in the proposed work is 0.0905 approximately. The average value
obtained for number of migration in the proposed system is 2. The average value obtained for SLA violation in the
proposed system is 0.8862.
6. References
1. Zhang, F., Liu, G., Fu, X., & Yahyapour, R. (2018). A Survey on Virtual Machine Migration: Challenges, Techniques
and Open Issues. IEEE Communications Surveys & Tutorials.
2. Beloglazov, A., Abawajy, J., & Buyya, R. (2012). Energy-aware resource allocation heuristics for efficient
management of data centers for cloud computing. Future generation computer systems, 28(5), 755-768.
3. Bari, M. F., Boutaba, R., Esteves, R., Granville, L. Z., Podlesny, M., Rabbani, M. G., ... & Zhani, M. F. (2013). Data center
network virtualization: A survey. IEEE Communications Surveys & Tutorials, 15(2), 909-928.
4. Buyya, R., Beloglazov, A., & Abawajy, J. (2010). Energy-efficient management of data center resources for cloud
computing: a vision, architectural elements, and open challenges. arXiv preprint arXiv:1006.0308.

5. Kapil, D., Pilli, E. S., & Joshi, R. C. (2013, February). Live virtual machine migration techniques: Survey and research
challenges. In Advance Computing Conference (IACC), 2013 IEEE 3rd International (pp. 963-969). IEEE.
6. Clark, C., Fraser, K., Hand, S., Hansen, J. G., Jul, E., Limpach, C., ... & Warfield, A. (2005, May). Live migration of
virtual machines. In Proceedings of the 2nd Conference on Symposium on Networked Systems Design &
Implementation-Volume 2 (pp. 273-286). USENIX Association.
7. Seth, S., & Singh, N. (2017, May). Dynamic Threshold-Based Dynamic Resource Allocation Using Multiple VM
Migration for Cloud Computing Systems. In International Conference on Information, Communication and
Computing Technology (pp. 106-116). Springer, Singapore.
8. Han, Y., Chan, J., Alpcan, T., & Leckie, C. (2017). Using virtual machine allocation policies to defend against co-
resident attacks in cloud computing. IEEE Transactions on Dependable and Secure Computing, 14(1), 95-108.
9. Ahmad, R. W., Gani, A., Hamid, S. H. A., Shiraz, M., Yousafzai, A., & Xia, F. (2015). A survey on virtual machine
migration and server consolidation frameworks for cloud data centers. Journal of Network and Computer
Applications, 52, 11-25.
10. Hameed, A., Khoshkbarforoushha, A., Ranjan, R., Jayaraman, P. P., Kolodziej, J., Balaji, P. & Khan, S. U. (2016). A
survey and taxonomy on energy efficient resource allocation techniques for cloud computing
systems. Computing, 98(7), 751-774.
11. Xu, F., Liu, F., Jin, H., & Vasilakos, A. V. (2014). Managing performance overhead of virtual machines in cloud
computing: A survey, state of the art, and future directions. Proceedings of the IEEE, 102(1), 11-31.
12. Radhakrishnan, A., & Kavitha, V. (2016). Energy conservation in cloud data centers by minimizing virtual
machines migration through artificial neural network. Computing, 98(11), 1185-1202.

IRJET- Research Paper on Energy-Aware Virtual Machine Migration for Cloud Computing

More Related Content

What's hot (20)

Similar to IRJET- Research Paper on Energy-Aware Virtual Machine Migration for Cloud Computing (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Research Paper on Energy-Aware Virtual Machine Migration for Cloud Computing