Skip to main content

Advances, Systems and Applications

Virtual machine placement in energy aware load balancer using fog classifier


Cloud datacenter carries huge volume of data and tasks which is allocating resources to multiple workstations. Most of the cloud services are operating service level agreement (SLA) placements. During execution datacenter emits carbon and makes the energy. So operation cost always consideration fact. We need to address this challenge by using energy aware load balancer. This load balancer can be fixed in Virtual Machines (VM) and Classifier is required for selecting VMs. Employing the VMs is very important factor so fog enabled services is required for distributed geo physical load balancer with energy efficiency. In this paper we propose offloading VM services and Fog classifier for load balancing the cloud services. Placing the VM from one host to another we use Host Load Balancer with Energy Aware placement algorithm. In this case dynamical cloud environment can be tested and compare the host results. This is empirical approach for place the VMs without compromising the users. The simulations are done by using CloudSim and TensorFlow is used of generating deep belief network model for preparing VM placement. Our proposed method achieves 96% energy efficiency with minimum migration cost. The results are compared with existing placement methods based on active host availability.


Cloud Services are very important factor to provide online based services such as software, application, platforms, etc. Virtual machine is selected based on request and infrastructure requirements from the user. Nowadays ubiquitous services are used for various computing applications [1]. Computing and storage services are playing extreme operations over the internet. The technological shirt is the major reason to enable convenient, on-demand and resource sharing operations. It is configurable computing paradigm with pay per usage model. Fog computing is the cloud edge level service to extend networking, storage and processing the clouds to the users at different resource utilization [2].

Clouds and Fogs are the employ of Virtual machine and the major focus for the VM is migration and replications. The above two factors are depended based on load balancing and energy efficiency. The major role is placing virtual machine and balancing the load in running conditions. Due to over machine maintenances and large volume of data processing work load for the VM is always in running stage. This factor is investigated by various literatures and we need to fix better relationship between VM workload [3] and number action users. In this case according to the result given Amazon web Service EC2 cloud [4] has number user varies means it has low value as 1% in intensive application and increases up to 60% CPU optimized load balancing conditions [1].

According the researchers point, various fog computing [5] problems and challenges are addressing in cloud applications. While implementing load balancer means we must consider two factors such as energy efficiency and placement of VMs [6]. In terms of Quality of Service we need provide for latency service providers and multiple users can be accessed via fog servers. The energy efficiency is obtained from fog server placements, profiles and applications. The number of downloading [7], access specification [8], service level agreements [9], number of updates [10] and amount data pre-loading [11] are considers for fixing VM with good energy efficiency factor.

In this paper, we give energy aware load balancer framework fixing VMs using Fog classifier. In this case we optimize the cloud with better energy efficiency factor. It is integrated linear programming model with minimize total energy consumption and multifactor VM workloads [12]. VMs workload can be measured by using number of user profiles and number nodes are in active stage. Using above two factors is considered for calculating data rate. Data rate is the major factor for fining energy efficiency once we achieve good and efficient energy means VM is placed in exact position [13]. This paper organize as follows, section 2 explains various related works, section 3 gives proposed classifier with energy optimization, section 4 experiments various methods and section 4 gives conclusion and future scope.

Related work

Data centers are the highest deliverable server to distribute the service across the world. For implementing cloud application a small number to 1000 server is required. Wong et al. large amount of energy is required when the server is increased [14]. The report is given by National Record Data Center US while setting server more than 90 billion kilowatt hour of electricity was used and every year it increased double the rate by 2021 annual report. So every year nearly 200 million tons of carbon pollution happened [15].

The geographical distributed data center sites have various cloud service providers such as Google, IBM, Microsoft and Amazon. Recently VMware provides good infrastructure computing resource sharing facility and multi user availability [16]. Manikandan et al. Placing Virtual machine is tedious process we need strong cohesion factor which addresses the problem for clustering, optimization and scheduling the resources. Cloud provider faces load balancing issue it affects overall VMs performance [17]. Yung et al. various selection processes is available for setting data center sites with respect to SLA and QoS factors. Each data center get power from different servers and it can be recorded by using on-site energy services [18].

The present system has coupling issue like cloud provider and cloud data center optimization. Normally both factors were affected by unbalancing scheduling and cloudlet from the VM [20]. Dynamically services can be accessed from data centers and migration is also reducing the load optimization. Xiang et al. utilization of servers and data center is always increased to automatically energy consumption is also increased. So we need good VM placement to handle the data centers and servers. The survey from iMatrix on 2022, some countries are increased the carbon tax if the industries emits the carbon while using servers and it affect environment sustainability [20].

George and Thimmana et al. monitoring the energy consumption and balancing the load is major issue so it is directly affect the entire server and it creates high consumption factor. Researchers are calculating different weight ratio which occupied by workload and physical resource availability. Based on ratio we consider below parameter while setting VMs using energy aware load balancer i. Calculate energy consumption factor and work load, ii. Find the minimum critical occupied resources, iii. Cross validation index with respect to VM and iv. Number of resources are in active stage. Above consideration are taken as important and we provide good VM placement with respect to energy consumption factor.

Cloud-fog model

In this work, Fog classifier is proposed to handle VMs placement based on type, programming model and linearity resources. The Fig. 1 shows the layer architecture model for setting cloud optimization. Here data transmission can be done by using IP over wide area network technology, core volume of servers used and connected. Passive optical network connection is established with optical line terminal and optical network unit for data transmission. Cloud data centers are connected with IP network and Fog classifier extends the service via edges. In this work we considered fog nodes are the access specifiers to handle the resource based on clusters.

Fig. 1
figure 1

Cloud data center with fog nodes

Based on above input the major problem with data centers is placement of VM. Most of computing it is not done properly due the active and dynamic energy utilization. So we need consider below input while designing Fog classifier.

  1. a.

    While placing VMs must check with sever whether it is fully utilized

  2. b.

    To minimize the cost of the energy in data centers we need to shutdown the servers when it is in idle stage.

  3. c.

    Minimize the energy consumption in the data centers and estimate the carbon tax of each resource allocation at physical infrastructures.

  4. d.

    Select the data centers based on availability and virtual provisioning

In this paper we proposed three VM placement methods are experimented by referring energy consumption and cost. Factor 1: Verify the physical resource factors such as power supply unit, cooling, UPS backup and lighting, Factor 2: Data center availability, Factor 3: Distributing load balancing unit among the data centers, Factor 4. Renewable energy resources.

From the above Fig. 2, VM placement is done in data centers with respect to all geographical areas considered overhead energy as less, optimized server and footprint of each resource allocation. VM placement can be done by using maximise the renewable energy utilization and minimum total cost. In this method we sort the entire host in descending order based on utilization factor. If utilization is less means we consider that host as weak and we used that for overloading conditions. This process can be repeated while fixing good placement of VM. Here we considered detecting under loaded hosts, host utilization factor, data center workloads and vacated VMs. Existing method also considered for targeting the host. In this case we propose Fog classifier for selecting VM and host optimization.

Fig. 2
figure 2

VM Placement and Selection based on availability

Fog classifier—host load balancer with energy aware placement algorithm

The prediction of number of host and fetch the lower threshold values. In this case fog classifier divides the host into lower threshold hosts and higher threshold hosts. So VM can select based on minimum changes based on host availability (Table 1).

Table 1 Fog classifier algorithm

Cloud users can submit the VM Request to cloud broker and select the VM values for the user so this case the time can be obtained as.

VMi = (Type of Service, Hold_Time) the type selected by Amazon EC2 VM instance and running out measured by using first come first serve basis.

Power Consumption in Data Center – Data center connects the power from UPS or Networking devices. This is distributed to all the data center components and uses the less energy consumption (Table 2). The cost of energy can be obtained as follows

$${\mathrm{PowerSupplyUnit}}_{\mathrm{Index}}\frac{Power(total)}{Power(utlized)}= \frac{Po-(Ps*Pu)}{Power(total)}$$

where Po – Overhead Power and Utilized power obtained from data centers.

Table 2 Conditions and task allocation in VMs


Experimental setup

Our proposed system is evaluated at Infrastructure as a service (Iaas) cloud environment using Amazon EC2 Cloud VM and CloudSim toolkit is used to test real time environments. It is large scale platform so evaluation can be done virtualized load balancer model. Here Fog classifier is added the features such as cluster the VMs, Host List, VM Allocation, Minimum power consumption factor and overhead values. The below Fig. 3 shows that VM Machine server with types.

Fig. 3
figure 3

VM Workloads at different types

VM – Types


VM – Memory

Bandwidth of VM


Instructions per seconds


512 MB

3 GB

100 mb/s




1 GB

3 GB

100 mb/s




2 GB

3 GB

100 mb/s




5 GB

3 GB

100 mb/s



We designed the data centers with Data center utilization, Available energy, Energy aware request from server and cloud information set. We are operating the data center we used hypervisor created by virtual machine with memory, bandwidth, CPU shown in Tables 3 and 4.

Table 3 Data center selection based on VM request from resource pool
Table 4 Energy aware consumption results after allocating VMs based on user requests

VMs placement

We used optimal placement using Data Center selection using core network topology consist of 30 nodes and 120 bidirectional link. In below Fig. 4 Core network topology from CISCO Network index representation. Here the 1300Gbps data rate can be applied with XGPON Model selected from VMs. Also more the 30 K user request sent to internet traffic and 60% user can active in dynamic stage. So the VM workload assigned as cluster group 1,4,8,16,25,50,100 and index set as 100%,50%,25%, 10%, 5% and 1%. The number of VM is considered as CPU capacity and active user status. The VM workload may varies based on CPU capacity and requirements. In this case data rate varies as 1Mpbs, 5Mpbs, 10Mpbs and 25Mpbs. For example lower to higher data rate can be assigned based on VM usage.

Fig. 4
figure 4

Core network topology VM allocation using active users

CloudSim experimental inputs

From above Table 5, experiments are set in cloudsim and apply our proposed algorithm for evaluating cloud. Here Fog classifier classify the cloud based on usage and availability. Proposed method is allocate maximum number of task to active host. The energy consumption is propotional to the action host in the dataset. If we increase the number of host means result will be decreased or shutdown the data center.

Table 5 Data center, server and VM characteristics for experiments

Based on above Table 6 and Fig. 5, VM Types tested by using various data center and number of host. In this case the energy consumption is below 35% and accuracy index average in 96% achieved. Also each stage number VMs are shutting down based on number of VM migrated. From this result our proposed algorithm is compared with existing VM placement and accuracy index values using TensorFlow. TensorFlow is the comparison tool to compare the results with various representations.

Table 6 Experimental result of VMs placement and accuracy index
Fig. 5
figure 5

Graphics representation of VM usage and accuracy

Based on above Table 7, the proposed method is compared with existing method virtualized load balancer and decision tree index methods with selecting VM Types, Data center and number of host. In this case comparison is taken as accuracy and number data centers are shutdown. From the two factors our proposed method Host Load Balancer with Energy Aware placement algorithm gives good accuracy and number of data centers also shut downed.

Table 7 Comparison of VM types with existing and proposed method


Virtual machine placement is major consideration in cloud computing environment. Because in cloud environment has more number of data centers, nodes, active user and dynamically changing resources. So we provided a solution to handle the cloud environment and balancing the load with less energy consumption using Fog classifier. We presented a framework with less energy consumption over a cloud-fog model. In this case VMs workloads are considered as major key player and simulate the cloud using CloudSim. Number hosts, Data Center and VM Types are taken and proposed Host Load Balancer with Energy Aware placement algorithm is applied to simulate the environment. From this our proposed system achieves 96% energy efficiency and it is compared with existing methods. In future may be our proposed algorithm is applied for hybrid or heterogeneous cloud environment.

Availability of data and materials

Not applicable.


  1. Manikandan S, Chinnadurai M (2022) Virtualized load balancer for hybrid cloud using genetic algorithm. Intell Autom Soft Comput 32(3):1459–1466

    Article  Google Scholar 

  2. Beloglazov A, Buyya R (2010) Adaptive threshold-based approach for energyefficient consolidation of virtual machines in cloud data centers. In ACM Proceedings of the 8th International Workshop on Middleware for Grids, Clouds and e-Science. Vol. 4

  3. Buyya R, Yeo CS, Venugopal S, Broberg J, Brandic I (2010) Cloud computing and emerging IT platforms: vision, hype, and reality for delivering computing as the 5th utility. Future Gener Comput Syst 25:599–616.

    Article  Google Scholar 

  4. Nathuji R, Schwan K (2007) Virtual power: coordinated power management in virtualized enterprise systems. ACM SIGOPS Oper Syst Rev ACM 2017(41):265–278

    Article  Google Scholar 

  5. Verma A, Ahuja P, Neogi A (2018) pMapper: power and migration cost aware application placement in virtualized systems. Proceedings of ACM/IFIP/USENIX 9th International Middleware Conference, Leuven, pp 243–264

    Google Scholar 

  6. Mell P, Grace T (2019) The NIST definition of cloud computing (draft). NIST Spec Publ 800:145

    Google Scholar 

  7. Patel N, Patel H (2020) Energy efficient strategy for placement of virtual machines selected from underloaded servers in compute Cloud. J King Saud Univ Comput Inform Sci 32:700–708

    Google Scholar 

  8. Kortas N, Youssef H (2019) Energy consumption of virtual machine migration within varied DCN architectures. IEEE, Conference Paper on 15th (IWCMC)

  9. S. Manikandan, P. Dhanalakshmi, K. C. Rajeswari and A. Delphin Carolina Rani (2022) Deep sentiment learning for measuring similarity recommendations in twitter data. Intell Autom Soft Comput 34(1);183–192

  10. Giacobbe M, Celesti A Fazio M, Villari M, Puliafito A (2015) Evaluating a cloud federation ecosystem to reduce carbon footprint by moving computational resources. Proc Symp Comput Commun 99–104

  11. Celesti A, Puliafito A, Tusa F, Villari M (2013) Energy sustainability in cooperating clouds. Proc 3rd Int Conf Cloud Comput Serv Sci 83–89

  12. Lin M, Wierman A, Andrew L, Thereska E (2011) Dynamic right-sizing for power-proportional data centers. Proc IEEE INFOCOM 1098–1106

  13. “Latest Microsoft Data Center Design Gets Close to Unity PUE,” [Online]. Available: Accessed on: 7 Apr 2017

  14. Wasim R et al (2015) A survey on virtual machine migration and server consolidation frameworks for cloud data centers. J Netw Comput Appl 52:11–25

    Article  Google Scholar 

  15. Dong X, El-Gorashi T, Elmirghani JMH (2011) Green IP over WDM networks with data centers. J Light Tech 29(12):1861–1880

    Article  Google Scholar 

  16. Al-Quzweeni AN, Lawey A, El-Gorashi TEH, Elmirghani JMH (2019) Optimized energy aware 5G network function virtualization. IEEE Access 7

  17. Musa MOI, El-Gorashi TEH, Elmirghani JMH (2018) Bounds on GreenTouch GreenMeter network energy efficiency. J Light Technol 36(23):5395–5405

    Article  Google Scholar 

  18. Cisco (2018) Cisco Global Cloud Index鳥: Forecast and Methodology , 2016–2021

    Google Scholar 

  19. Google Cloud, “Cloud locations.” [Online]. Available:

  20. Amazon, “London in the Cloud.” [Online]. Available:

  21. Netflix (2019) Internet Connection Speed Recommendations. Netflix Help Center

Download references


Not applicable.


Not applicable.

Author information

Authors and Affiliations



Mr. S. Selvaganapathy wrote the main manuscript text and Dr. M. Chinnadurai reviewed the manuscript.

Corresponding author

Correspondence to S. Selvaganapathy.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

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

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Selvaganapathy, S., Chinnadurai, M. Virtual machine placement in energy aware load balancer using fog classifier. J Cloud Comp 12, 180 (2023).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: