Introduction

Nowadays Cloud Computing is used in all the business environments such as multinational companies, industries, banking sectors and other commodity environments. Day by day usage of resources and applications are also increased1. The high demand of availability and resource request is evolving in services. Current scenario cloud provides anything as a Service such as infrastructure, platform and application as a services2.

The cloud services are internet based service offerings and many countries are using large volume of resources and resource computations. In this case allocation of VMs and optimizing resources are major issues3. According to the recent cloud survey, virtualization is used in all the cloud services. Which share the resources to multiple users based availability and resource request. Major concern is handling resource with less energy consumption. In this case VMs are mapped with cloud data center so we can access the physical machine are easily.

The cloud services are used in all kind of online services during and after pandemic the usage of resources are day by day increased4. The volumes of resources are stored in cloud space based on SLAs and VMs resources are utilized. The Cloud resources are classified based on their storage level, access controls and turnaround time of each request. In this case each request is classified based on cloud space and cloud service provider resources. Each service is demonstrated based on their consolidated data center features and each physical storage. When we are using Hybrid cloud environment the volume of physical data are recorded and problems are addressed5.

Hybrid Cloud Environment has huge volume of physical machines which composed as logical machines such as VM consolidations6. There are several user can access the resource so we need to provide less energy consumed resource with VM consolidation and better accuracy results. In this paper, “Related works” section deals various related works about VM consolidation and energy utilization, “Hybrid cloud optimization—VM consolidation” section gives problem statement with proposed methods, “VM consolidation—experiments” section explains proposed algorithms with simulations, “Experimental setup” section gives experiments and conclusion.

Related works

The result 2023 cloud magazine the two platforms such as Azure and Google provided less CPU utilization as less than 25%7 and IBM has average result in between 25 and 30%8. Same way Amazon Web Services has more utilization factor and more resources can be accessed by variety of users9. The CPU utilization and energy consumption is major consideration to give effective resource access services. The below Table 1 shows that various cloud services utilization factors.

Table 1 Cloud Platform result of energy consumption based on Cloud Magazine (2023).
Full size table

The Consolidation of VMs are tedious process in cloud environment because we need to considered utilization of resources10, execution11 and turnaround time of CPU12, service level agreements13 and service offerings14. While implementing this we need to take care over and underutilization of resources such migration, placement and optimization. According various related studies need to take care constraint stratification problem, genetic methods, heuristics solutions and NP-hard completeness solutions. Virtualizing data centres is done by using linear programming model based on heuristics solutions. Our proposed idea is to generate decision tree model and trace the parameter from reference or backtracked results15.

Many researchers are suggested that VM allocation in Cloud is different scenarios. Lion optimizer is used to allocate VM on Cloud data center with physical features. From this related research studies we need an efficient VM consolidation method with reduced power consumption, reduced resource wastages and optimal allocations.

Hybrid cloud optimization—VM consolidation

Cloud Optimization is the strategies approach and constructed based on linearity, evolutionary and constraint approaches. In this paper we propose adaptive backtracking VM consolidation approach to optimize the resource with lesser energy consumption. Here we use Adaptive Hill Climbing method to generate hybrid cloud with backtracking resource features using deep learning and Pursuit algorithm is used to calculated utilization factors. The below factors are considered to implement our system,

  1. a.

    Generate VM consolidation using adaptive backtracking with consideration of SLA violations, VMs, energy factors and migration.

  2. b.

    Design a VM migration tool and to select under and over utilized machine based on real time movement and threshold values.

  3. c.

    Place VM based on the request and availability which has generation strategy, crossover issues and development issues.

  4. d.

    Simulate the experiments using CloudSim for generating cloud platforms and Matlab is used to measure the accuracy and results.

Figure 1 show that assigning resources to VM based on migration result. The live VM migration is applied to move underutilized physical resources to sleep mode and utilized or running machines in active node. So we can consolidate the VMs based on this factor. In this case SLA violation is occurred to negative impact will be created and delay the response time of the user application. Live Migration is addressed by using migrate VMs to overloaded physical machines based on capacities. This process is called VM placement consolidation and it is decision making category of each resources.

Fig. 1
figure 1

VM placement, management and configuration.

Full size image

Virtual Machine (VM) is presented as

$$ {\text{Machine}}_{{\text{V}}} = \left\{ {\begin{array}{*{20}l} 0 \hfill & {{\text{i}} = {\text{NULL}}} \hfill \\ 1 \hfill & {{\text{if i }} = {\text{ HOST at Machine}}_{{\text{P}}} } \hfill \\ \end{array} } \right., $$
(1)

where Machine has two options such as physical and virtual based on host and i is represented as number of resource availability.

$$ {\text{Machine}}_{{\text{p}}} = \left\{ {\begin{array}{*{20}l} 0 \hfill & {{\text{i}} = {\text{NULL}}} \hfill \\ 1 \hfill & {{\text{if i }} = {\text{ HOST at Machine}}_{{\text{v}}} = {\text{Available}}} \hfill \\ \end{array} } \right.. $$
(2)

Resource Optimization and Utilization are the major challenges to be selected based on cloud service provider availability. From the above Table 2 (Algorithm 1) we migrate the VMs from underutilized status to over utilized status. It is classification process and it is done by using VM migration polices as live status. So the negative system performance is reduced. Finally we get balanced migrated resource sharing system.

Table 2 Hill climbing algorithm to process the resources.
Full size table

VM consolidation—experiments

The physical machine resources are utilization can be done by using dynamic load balancing systems. In the scenario we can set the threshold values based manual operations. The below algorithms is shows that VM consolidation, migration and resource utilization index measurements in Table 3.

Table 3 Pursuit algorithm—calculated threshold values based on VM placement.
Full size table

The threshold challenges is addressed based on moving ranged resource and it is automatically calculated the resource utilization index. The physical machine resources obtained based on cloud provider results and threshold results. Moving range is measured by using continuous data analytics features from underutilized factors. So we can dynamically allocate the resources regarding CPU utilization, VM migrated results, lower bound values and VM consolidation features.

Figure 2 shows that VM consolidation based on encoding schemes with PM and VM features. It is allocated based on resource availability and utilization factors.

Fig. 2
figure 2

Flowchart—VM placement and consolidation.

Full size image

The VM placement solution is encoded in the form of chromosomes. Each chromosome is represented as a gene order, with each gene representing a PM where the VM will be placed. Table 4 shows an example of Virtual machines placed to physical machines and their equivalent chromosome when the number of visual machines is 10 and the number of physical machines is 4. The V1, V2, V3 are placed in M1, while V4, V5 are placed in PM2, and V6, V7, V8, V9 are placed in PM3 finally V10 is placed in PM4.

Table 4 Virtual and physical machine classification based resource utilization.
Full size table

The pursuit algorithm is implemented based on specific VM placement setting up lower threshold values. The higher threshold values are classified as over utilized factor VM consolidated index is recorded based on following Fig. 2 flowchart. This case we first select the parent node with lowest threshold values and identify the location. Second place the VM based on second node representations in Table 5.

Table 5 Classify the physical machine based VM placement consolidation.
Full size table

From the Fig. 2, VM consolidation can be done based on threshold values and multiple resource utilization. The fitness function is calculated based number over and under utilized threshold values and physical machine representations. The following is fitness function of utilized resource with x population.

$$ {\text{MinFF}}\left( {\text{x}} \right) \, = {\text{ OverUtil }}\left( {\text{X}} \right) \, + {\text{UnderUtil }}\left( {\text{X}} \right). $$
(3)

From the above algorithm VM consolidation can be done by using hill climbed results and then pursuit algorithm used to simulate the VM placement and migrations. Above algorithms are tested by using experimental setup in Table 6.

Table 6 VM consolidation using threshold features.
Full size table

Experimental setup

The Adaptive Backtracking method is simulated by using Matlab environment. Hybrid Cloud Environment created by using CloudSim and Computing system has Intel Core i7 GPU Computer, 16 GB RAM, 1 TB HDD and Window 1O OPS. The performance factors are calculated as follows.

Energy consumption factor

The energy efficiency is calculated based on Physical machine resource utilized factor values and number of VM placed,

$$ {\text{ECF }}\left( {{\text{Machine}}_{{\text{P}}} } \right) \, = {\text{ Machine}}_{{\text{P}}}^{{{\text{Idle}}}} + {\text{ Machine}}_{{\text{P}}}^{{{\text{Full}}}} + {\text{ VM}}\left( {\text{n}} \right) \times {\text{Threshold}}. $$
(4)

VM Migration is calculated from using number over and underutilized resource and availability factor. The fitness function gives the results for VM values status. CPU utilization is calculated based on migrated VM results which pull down the performance and execution time. The accuracy factor is measured by using turnaround time of each executed results.

$$ {\text{Accuracy}}\left( {\text{i}} \right) \, = {\text{i}} = {\text{1NVMAvailable}} + {\text{VMEmptyMachineP,}} $$
(5)
$$ {\text{SLAviolation }} = {\text{ Accuracy}} \times {\text{ECF}}. $$
(6)

From the above experiments below Table 7 shows that the result of accuracy and power consumption results based on number of VMs.

Table 7 Experimental result of VMs placement and accuracy factory.
Full size table

The result from above Table 7 the average accuracy in 96% and Energy efficiency is achieved as 0.22. This is very less energy consumed while increased VMs and physical resources in Fig. 3. The above results are compared with existing methods and shown in Table 8.

Fig. 3
figure 3

Average VM consolidation and energy consumption based on iteration results.

Full size image
Table 8 Comparison of existing methods with proposed system—accuracy and energy constraints.
Full size table

This case our proposed method results are compared with existing methods such as Decision tree, K-means, Support vector and TensorFlow graph. The results are compared average cpu utilization, turnaround time and energy efficiency index in Fig. 4. The results show that our proposed methods achieved better results.

Fig. 4
figure 4

Comparison of turnaround time and energy efficiency index.

Full size image

In this paper we compared the execusion results with some existing VM Scheduling and consolidation methods. The below Table 9 shows that the comparison representation of our proposed method CPU utilization and energy efficiency factors.

Table 9 Comparison result of recent research methods and our proposed method.
Full size table

From the above Table 9 results, the current research methodologies in cloud computing such as Dynamic Virtual Machine Scheduling Using Residual Optimum Power-Efficiency In The Cloud Data Center, SR-PSO: server residual efficiency-aware particle swarm optimization for dynamic virtual machine scheduling, VMS-MCSA: virtual machine scheduling using a modified clonal selection algorithm results are compared with our proposed method Adaptive Backtracking. Based on the results our proposed method gives 96% accuracy factor (CPU Utilizations) and lesser energy efficiency index.

Conclusion

Cloud computing is the major source to provide various services based on user request. Virtualization is another important factor to convert physical resources in logical services like VMs. We proposed adaptive backtracking VM consolidation method to provide lesser energy consumption and better accuracy results. Hill climbing algorithm is applied to classify the VMs from physical machine resource and availability. Pursuit algorithm is applied to measure threshold values to found over and underutilized resources. From the results Matlab simulator is used to simulate the systems and calculated accuracy as 96%. Also we achieved efficient energy consumption and compared the results with existing methods. In future same consolidation factor can be proposed for live VMs.