Machine Learning to Estimate Workload and Balance Resources with Live Migration and VM Placement

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper "A Hybrid Machine Learning Model to Estimate the Live Migration of Virtual Machine Placement" presents an interesting approach to optimizing VM placement using machine learning techniques. However, the paper lacks scientific rigor in its presentation of results. There is a notable absence of visual representations, such as graphs and confusion matrices, which are essential for clearly demonstrating the model's performance.

Few points to improve the paper.

The abstract is too short. Enhance it by emphasizing the crucial role of the proposed study and clearly outlining the authors' contributions in terms of performance enhancements and innovation. This will provide readers with a better understanding of the research findings' significance and the authors' unique contributions.

 The Introduction section requires further elaboration, specifically focusing on Live Migration of Virtual Machine Placement. The entire introduction should center on explaining the study's focus rather than providing a general introduction to the problem and key terms used in the paper.   

 In the Methods section, the objective of the research should not be reflected; it should be placed at the end of the Introduction section.

 Most of the figures in this research are taken from different researchers, which shows the reliance on existing literature and the need to contextualize findings within the broader research landscape with authors contributions.

 To address the issues in the results section for 'A Hybrid Machine Learning Model to Estimate the Live Migration of Virtual Machine Placement,' where claims are not shown through graphs and confusion matrices are missing, and to discuss the authenticity of the results, how the results are validated, and how the objectives are achieved. The authors mention 99% accuracy through Hybrid Machine Learning MDP + GA + RF but missing in the result section.

Comments on the Quality of English Language

English is ok

Author Response

Reviewer#1, Concern # 1:

The paper "A Hybrid Machine Learning Model to Estimate the Live Migration of Virtual Machine Placement" presents an interesting approach to optimizing VM placement using machine learning techniques. However, the paper lacks scientific rigor in its presentation of results. There is a notable absence of visual representations, such as graphs and confusion matrices, which are essential for clearly demonstrating the model's performance.

Few points to improve the paper.

Author response:  Thank you for the thoughtful question.

Reviewer#1, Concern # 2:

The abstract is too short. Enhance it by emphasizing the crucial role of the proposed study and clearly outlining the authors' contributions in terms of performance enhancements and innovation. This will provide readers with a better understanding of the research findings' significance and the authors' unique contributions.

Author response:  Thank you for the valuable suggestion.

Author action: We have revised the paragraph on abstract:  The paper presents a hybrid machine-learning model for estimating the pre-copy live migration of virtual machines in a data center. The proposed model combines of Markov Decision Process (MDP), Genetic Algorithm (GA), and Random Forest (RF) algorithms to predict which virtual machines will be moved first and determine the best host machine target. The hybrid model achieved 99% accuracy with faster training times compared to previous studies using K-Nearest Neighbor, Decision Tree Classification, Support Vector Machines, Logistic Regression, and Neural Networks. The authors suggest further research using a deep learning approach (DL) to solve other issues related to data center performance. The article presents a promising strategy for improving virtual machine migration in data centers. The hybrid model demonstrates high accuracy and faster training times compared to previous studies, indicating its potential to optimize virtual machine placement and reduce downtime. The authors also highlight the importance of considering data center performance and suggest further research. Additionally, it would be useful to see more information on the practical implementation and deployment of the proposed model in real-world data centers.

Reviewer#1, Concern # 3:

The Introduction section requires further elaboration, specifically focusing on Live Migration of Virtual Machine Placement. The entire introduction should center on explaining the study's focus rather than providing a general introduction to the problem and key terms used in the paper.   

Author response:  Thank you for the valuable suggestion.

Author action: We have revised the paragraph Introduction on page 2, In the introduction I have presented research contributions:

• The authors propose a novel hybrid machine learning model that combines Markov Decision Process (MDP), Genetic Algorithm (GA), and Random Forest (RF) algorithms to estimate the pre-copy live migration of virtual machines (VMs) in a data center.
• The proposed model can predict which VMs will be moved first and determine the best host machine target to optimize VM placement and reduce downtime.
• The authors compare the proposed model with existing state-of-the-art machine learning algorithms, such as K-Nearest Neighbor, Decision Tree Classification, Support Vector Machines, Logistic Regression, and Neural Networks. The results show that the proposed model achieves 99% accuracy with faster training times than previous studies.
• The authors suggest further research using a deep learning approach (DL) to solve other issues related to data center performance.

Improve the accuracy of the learning machine in the live migration technology decision-making processes, our research goal is to develop rules for LVM that focus on improving the accuracy of the machine learning model and balancing DC workloads. To achieve this, we consider the balance of the workload of the VM as a key factor. The balance of VM workloads is determined by CPU and memory, as well as by which LVM scheduling is prioritized for migration.

Reviewer#1, Concern # 4:

In the Methods section, the objective of the research should not be reflected; it should be placed at the end of the Introduction section.

Author response:  Thank you for the valuable suggestion.

Author action: We have revised the paragraph on page 7 (Section III):

3. Methods

In this section, we outline the objectives of our research and then define the problem, can be seen in Figure 3 as follows:

 

Figure 3. Flowchart of Methodology Research

Figure 3 is a flowchart that outlines the research process for the study on live virtual machine (LVM) migration using a hybrid machine learning approach. The process is divided into five main stages: introduction, literature review, data collection, model development, conclusion, and future work. The introduction stage defines the problem and motivation for the study, provides background on LVM and machine learning, and states the research objectives and questions. The literature review stage summarizes previous LVM and machine learning studies, identifies gaps and limitations in current research, and formulates hypotheses for the study. The data collection stage describes the data sources and collection methods, explains data preprocessing and cleaning, and presents the final dataset used in the study. The model development stage introduces the evaluation metrics used to assess the model's performance, compares the hybrid machine learning approach results with previous studies, and discusses the model's limitations and potential improvements. Finally, the Conclusion and Future Work stage summarizes the study's main findings and contributions, discusses the study's implications for data center performance and management, and suggests directions for future research on LVM and machine learning in data centers. Overall, Figure 3 provides a clear and concise visual representation of the research process for the study, highlighting the key stages and their corresponding activities.

Reviewer#1, Concern # 5:

Most of the figures in this research are taken from different researchers, which shows the reliance on existing literature and the need to contextualize findings within the broader research landscape with authors contributions.

Author response:  Thank you for the valuable suggestion.

Author action: We have revised the paragraph on page 2 (Section II):

2.1. Previous studies of live migration technology

The authors discovered an intriguing gap in LVM improvement conducted by Haris et al. (2023) in ML, but we chose to enhance ML modeling with a more hybrid approach, making the LVM approach more optimal based on MDP and GA algorithms in decision-making and selection of LVM scheduling. This aligns with the 2020 research by Guo et al., arguing that the Markov Decision Process (MDP) is the type of NLP frequently used in reinforcement learning to provide the best policy [18]. However, the upcoming research aims to go further by developing hybrid machine learning to determine VM workloads with higher loads and presenting detailed information on LVM preparation. It involves selecting which VM instances should be migrated and when the optimal time for LVM execution is guided by optimal HM objectives. In this study, the MDP algorithm aligns with the research conducted by Alqarni et al. (2023), where problem-solving involves the reallocation of resources [30]. This hybrid ML can significantly have an impact on the prediction of overloaded VM issues and selection in scheduling processes in the LVM. In Table 1, we explore previous research conducted on LVM processes.

 

Table 1. Deploying Virtual Machines in Live Migration

Authors	Research Methods	Workload	Prediction targets	Key Findings
[12]	The pre-copy method	CPU	Memory Content Migration	There were no selection techniques for VM migration based on workloads.
[30]	Re-initialization and Decomposition-Whale Optimization Algorithm	CPU, Memory	Resource utilization	Only energy usage and migration costs; no specific virtual machine workloads were used.
[31]	Support Vector Regression	Memory, bandwidth, and CPU	Host utilization	It was limited to evaluating host usage; it was not suitable for evaluating the performance of VM migrations, although its precision was higher than that of other models.
[32]	Machine learning-based downtime optimization (MLDOM)	CPU, memory, and network utilization	Downtime	This process was conducted to reduce the downtime only applied to VM migration over the network environment.
[33]	Artificial neural network	large number of factors	Bandwidth usage and CPU Utilization	Performance prediction for pre-copy migration was not affected by increasing data center efficiency.
[34]	Genetic Algorithm	CPU, Memory, Network	CPU Utilization and Bandwidth	Fast VM placement was implemented.
[35]	Forecasting technique	Disk, Memory, and CPU	Disk, Memory, and CPU	Workload prediction was obtained to minimize VM migration.
[36]	Machine learning	CPU	CPU utilization	The k-nearest neighbor (KNN) is the approach and method in the decision of tree classification algorithms, which were compared to determine the number of VMs migrating with an accuracy level of 90.9%.

Reviewer#1, Concern # 6:

To address the issues in the results section for 'A Hybrid Machine Learning Model to Estimate the Live Migration of Virtual Machine Placement,' where claims are not shown through graphs and confusion matrices are missing, and to discuss the authenticity of the results, how the results are validated, and how the objectives are achieved. The authors mention 99% accuracy through Hybrid Machine Learning MDP + GA + RF but missing in the result section.

Author response:  Thank you for the valuable suggestion.

Author action: We have revised the paragraph on page 9-13 (Section IV):

4. Results

The Live Migration of VMs is upheld in several configuration scenarios, depending on which assets need to be relocated. The arrangement of relocation forms also has distinctive components that determine the type of movement. In this area, we examine in detail the setup and sorts of live relocation, can be seen in Figure 4 as follows:

 

 

Figure 4. Scenario Live Migration Model Configuration

Figure 4 illustrates the configuration of a live migration model for a virtual machine (VM) in a cloud computing environment. The model consists of three main components: the source host, the target host, and the VM being migrated. The source host is the physical server where the VM is currently running. It contains the VM's memory, CPU, and storage resources. In the live migration process, the source host continues to run the VM until the migration is complete. The target host is the physical server where the VM will be migrated to. It has the same hardware specifications as the source host. The target host is responsible for receiving the VM's memory, CPU, and storage resources from the source host and starting the VM on its hardware. The VM being migrated is the virtual machine that is running on the source host and will be moved to the target host. The VM's memory, CPU, and storage resources are copied from the source host to the target host during the migration process. The live migration process begins with the creation of a snapshot of the VM's memory, CPU, and storage resources on the source host. This snapshot is then transferred to the target host over the network. As the snapshot is being transferred, the VM continues to run on the source host. Once the snapshot is fully transferred to the target host, the VM's CPU and memory resources are stopped on the source host and restarted on the target host. The VM's storage resources are then synchronized between the source and target hosts to ensure data consistency. Finally, the VM is restarted on the target host and the migration process is complete. Overall, the live migration model configuration shown in Figure 3 enables the seamless migration of a VM from one physical server to another without any downtime or disruption to the VM's applications or services.

Table 5. Performance of the proposed hybrid model (%)

Model	Accuracy	Precision	Recall	F1 score
MDP + GA + RF	99	99	96	98

 

The proposed combined three algorithms: hyperparameters combined from MDP, GA, and RF. Table 6 shows the hybrid combination algorithm as in Table 5. Combining the MDP, GA, and RF algorithms offers several advantages. It is an algorithm that can take advantage of the strengths and weaknesses of other algorithms. MDP can model the probability between features and target variables, GA can provide the best features, and RF can make decisions on the selection of the best features. In implementing this hybrid ML as a whole, we can find which VM has the most workload on the HM and determine the right time for the LVM process to allow DC performance to be in a constant good condition. We use the following example in Figure 5 to illustrate the definition of the VM coverage to which HM.

 

 

Figure 5. Scenario Test 1 VM Placement and Migration (HM1, HM2, HM3)

Figure 4 shows a test scenario for VM placement and migration using the GWA-T-12 Bitbrains dataset [60, 61]. In this scenario, a migration process is carried out for HMs that have overload, using datasets from Bitbrains clouds. The availability of resources at the destination HM, including CPU, memory, and disk space, is taken into account when determining the placement of the VM. Additionally, equipment compatibility between the source and destination is considered to ensure that the VM can run properly after migration. Figure 6 shows the results of processing the dataset for this test scenario. The loading process on GA takes a long time, so only dataset resources of 156 MB are used. The live migration network is not carried out due to the need for network simulation via a router and the internet.

4.1. The result of the virtual machine of the dataset

 

Figure 6. result of the virtual machine of the dataset

In Figure 6, we tried to test the data set that had been created and then tested it with a hybrid ML algorithm. The results show that hybrid machine learning is capable of providing workload detection from virtual machines, in which hosts experience a greater workload. In the test dataset, we took samples from 30 virtual machines whose allocations were spread over HM 1, HM 2, and HM 3 and consisted of 284234 records. In Table 6 and Table 7, we test the algorithm, in which in Testing 1 we predict the workload on the HM that will experience migration using the MDP approach, and in Table 6 we determine the workload of the HM on the VM that will experience LVM.

Table 6. Results of dataset testing 1

HM allocation	Status
	Migration of virtual machines	No migration of VM	Total
HM 1	4901	85739	90640
HM 2	28942	78402	107344
HM 3	71	86190	86261

 

After predicting the workload on HM, we carried out tests using GA and RF algorithms with the variable workload of the VM and the network traffic to allow the LVM process to run optimally. The results can be seen in the data from test 2, in which HM 1, HM 2, and HM 3 have experienced changes, indicating that the workload in HM has decreased. The test results for Test 1 and Test 2 depend on the condition of the available VM workload.

Table 7. Results of data set testing 2

HM allocation	Status
	Predicted Migration Decision	No migration of VM	Total
HM 1	4648	85992	90640
HM 2	28752	78592	107344
HM 3	60	86190	86250

 

Figure 7. Results in Test 1 and Test 2

The results of the test in tests 1 and test 2 can be seen in Figure 7. The algorithm developed runs optimally.

5. Discussion

The discussion section of the paper provides a detailed analysis of the results obtained from testing the development in LVM process placement. The authors performed several tests and algorithm comparisons before combining the developed algorithms. They found that the KNN algorithm had an accuracy of 92% and 91% with longer training times, while the DT algorithm and logistic regression had faster training times but lower accuracy values of 91% and 89%. On the other hand, the RF algorithm had a high accuracy of around 93% and faster training time compared to the SVM algorithm. Therefore, the authors decided to combine the RF algorithm with MDP and GA, resulting in a precision value of 99% and an F1 score of 98%, indicating that the hybrid machine learning developed was more optimal than other algorithms in the case of LVM. Table 8 summarizes the results of the algorithm ML Live Migration VM, showing that the proposed MDP + GA + RF algorithm achieved the highest accuracy of 99.00% in the Live migration of VM placement scenario, outperforming other algorithms such as KNN, DT, logistic regression, and SVM. Overall, the results of the tests conducted in the study prove that the proposed hybrid machine learning algorithm is superior to past research, achieving high accuracy and faster training times in the LVM process placement scenario, , can be seen in table 8 as follows:

Table 8. Results Algorithm ML Live Migration VM

Authors	Algorithm of Machine Learning	Allocation Scenario Test	Result Test Accuracy (%)
This paper follows	MDP + GA + RF	Live migration of VM placement	99.00

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The authors are presenting an interesting paper regarding ML techniques to decide when and where to move a virtual machine from a host machine to another destination. However, the paper would benefit from the following observations:

- the authors did not mention anything about the data sets used in training the ML: the source of the data sets or if the same data sets were used by the articles used as comparison.

- the authors have mentioned CPU and Memory as features used in deciding the movement of a machine, but why not also power consumption?

- There is no analysis showing how changing the location of the VM had an impact on the HM resources. Are the results only relying on the data sets? If that is the case, then more details of the data sets should be provided and as mentioned if the same training and verification sets were used in the comparison papers.

- It is not mentioned how much it takes to train the hybrid ML (the proposed method).

- For some of the references is not mentioned or not so clear the year of publication (e.g., 10, 11, 16 etc.)

Author Response

Reviewer#2, Concern # 1:

The authors are presenting an interesting paper regarding ML techniques to decide when and where to move a virtual machine from a host machine to another destination. However, the paper would benefit from the following observations:

Author response:  Thank you for the support.

Reviewer#2, Concern # 2:

the authors did not mention anything about the data sets used in training the ML: the source of the data sets or if the same data sets were used by the articles used as comparison.

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 11 and 13 (Section III):

 

 

Figure 5. Scenario Test 1 VM Placement and Migration (HM1, HM2, HM3)

Figure 4 shows a test scenario for VM placement and migration using the GWA-T-12 Bitbrains dataset [60, 61]. In this scenario, a migration process is carried out for HMs that have overload, using datasets from Bitbrains clouds. The availability of resources at the destination HM, including CPU, memory, and disk space, is taken into account when determining the placement of the VM. Additionally, equipment compatibility between the source and destination is considered to ensure that the VM can run properly after migration. Figure 6 shows the results of processing the dataset for this test scenario. The loading process on GA takes a long time, so only dataset resources of 156 MB are used. The live migration network is not carried out due to the need for network simulation via a router and the internet.

 

Data Availability Statement: We used Bitbrains clouds datasets are available online at http://gwa.ewi.tudelft.nl/datasets/gwa-t-12-bitbrains [60, 61] .

 

Dataset has been used in articles:

Statistical Characterization of Business-Critical Workloads Hosted in Cloud Datacenters	Machine Learning	IEEE	https://ieeexplore.ieee.org/document/7152512
Dynamic Virtual Machine Consolidation Algorithm Based on Balancing Energy Consumption and Quality of Service	Machine Learning	IEEE	https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9843955
Datacenter Workload Classification and Characterization: An Empirical Approach	Machine Learning	IEEE	https://ieeexplore.ieee.org/document/8721402

 

Reviewer#2, Concern # 3:

The authors have mentioned CPU and Memory as features used in deciding the movement of a machine, but why not also power consumption?

Author response:  Thank you for the valuable suggestion.

Author action: Nowday, we focus on Machine Learning, for future research, we may discuss energy consumption in optimizing energy usage on VMs.

Reviewer#2, Concern # 4:

There is no analysis showing how changing the location of the VM had an impact on the HM resources. Are the results only relying on the data sets? If that is the case, then more details of the data sets should be provided and as mentioned if the same training and verification sets were used in the comparison papers.

Author response:  Thank you for the valuable suggestion.

Author action: We have revised the paragraph on page 9-13 (Section IV):

4. Results

The Live Migration of VMs is upheld in several configuration scenarios, depending on which assets need to be relocated. The arrangement of relocation forms also has distinctive components that determine the type of movement. In this area, we examine in detail the setup and sorts of live relocation, can be seen in Figure 4 as follows:

 

 

Figure 4. Scenario Live Migration Model Configuration

Figure 4 illustrates the configuration of a live migration model for a virtual machine (VM) in a cloud computing environment. The model consists of three main components: the source host, the target host, and the VM being migrated. The source host is the physical server where the VM is currently running. It contains the VM's memory, CPU, and storage resources. In the live migration process, the source host continues to run the VM until the migration is complete. The target host is the physical server where the VM will be migrated to. It has the same hardware specifications as the source host. The target host is responsible for receiving the VM's memory, CPU, and storage resources from the source host and starting the VM on its hardware. The VM being migrated is the virtual machine that is running on the source host and will be moved to the target host. The VM's memory, CPU, and storage resources are copied from the source host to the target host during the migration process. The live migration process begins with the creation of a snapshot of the VM's memory, CPU, and storage resources on the source host. This snapshot is then transferred to the target host over the network. As the snapshot is being transferred, the VM continues to run on the source host. Once the snapshot is fully transferred to the target host, the VM's CPU and memory resources are stopped on the source host and restarted on the target host. The VM's storage resources are then synchronized between the source and target hosts to ensure data consistency. Finally, the VM is restarted on the target host and the migration process is complete. Overall, the live migration model configuration shown in Figure 3 enables the seamless migration of a VM from one physical server to another without any downtime or disruption to the VM's applications or services.

Table 5. Performance of the proposed hybrid model (%)

Model	Accuracy	Precision	Recall	F1 score
MDP + GA + RF	99	99	96	98

 

The proposed combined three algorithms: hyperparameters combined from MDP, GA, and RF. Table 6 shows the hybrid combination algorithm as in Table 5. Combining the MDP, GA, and RF algorithms offers several advantages. It is an algorithm that can take advantage of the strengths and weaknesses of other algorithms. MDP can model the probability between features and target variables, GA can provide the best features, and RF can make decisions on the selection of the best features. In implementing this hybrid ML as a whole, we can find which VM has the most workload on the HM and determine the right time for the LVM process to allow DC performance to be in a constant good condition. We use the following example in Figure 5 to illustrate the definition of the VM coverage to which HM.

 

 

Figure 5. Scenario Test 1 VM Placement and Migration (HM1, HM2, HM3)

Figure 4 shows a test scenario for VM placement and migration using the GWA-T-12 Bitbrains dataset [60, 61]. In this scenario, a migration process is carried out for HMs that have overload, using datasets from Bitbrains clouds. The availability of resources at the destination HM, including CPU, memory, and disk space, is taken into account when determining the placement of the VM. Additionally, equipment compatibility between the source and destination is considered to ensure that the VM can run properly after migration. Figure 6 shows the results of processing the dataset for this test scenario. The loading process on GA takes a long time, so only dataset resources of 156 MB are used. The live migration network is not carried out due to the need for network simulation via a router and the internet.

4.1. The result of the virtual machine of the dataset

 

Figure 6. result of the virtual machine of the dataset

In Figure 6, we tried to test the data set that had been created and then tested it with a hybrid ML algorithm. The results show that hybrid machine learning is capable of providing workload detection from virtual machines, in which hosts experience a greater workload. In the test dataset, we took samples from 30 virtual machines whose allocations were spread over HM 1, HM 2, and HM 3 and consisted of 284234 records. In Table 6 and Table 7, we test the algorithm, in which in Testing 1 we predict the workload on the HM that will experience migration using the MDP approach, and in Table 6 we determine the workload of the HM on the VM that will experience LVM.

Table 6. Results of dataset testing 1

HM allocation	Status
	Migration of virtual machines	No migration of VM	Total
HM 1	4901	85739	90640
HM 2	28942	78402	107344
HM 3	71	86190	86261

 

After predicting the workload on HM, we carried out tests using GA and RF algorithms with the variable workload of the VM and the network traffic to allow the LVM process to run optimally. The results can be seen in the data from test 2, in which HM 1, HM 2, and HM 3 have experienced changes, indicating that the workload in HM has decreased. The test results for Test 1 and Test 2 depend on the condition of the available VM workload.

Table 7. Results of data set testing 2

HM allocation	Status
	Predicted Migration Decision	No migration of VM	Total
HM 1	4648	85992	90640
HM 2	28752	78592	107344
HM 3	60	86190	86250

 

Figure 7. Results in Test 1 and Test 2

The results of the test in tests 1 and test 2 can be seen in Figure 7. The algorithm developed runs optimally.

5. Discussion

The discussion section of the paper provides a detailed analysis of the results obtained from testing the development in LVM process placement. The authors performed several tests and algorithm comparisons before combining the developed algorithms. They found that the KNN algorithm had an accuracy of 92% and 91% with longer training times, while the DT algorithm and logistic regression had faster training times but lower accuracy values of 91% and 89%. On the other hand, the RF algorithm had a high accuracy of around 93% and faster training time compared to the SVM algorithm. Therefore, the authors decided to combine the RF algorithm with MDP and GA, resulting in a precision value of 99% and an F1 score of 98%, indicating that the hybrid machine learning developed was more optimal than other algorithms in the case of LVM. Table 8 summarizes the results of the algorithm ML Live Migration VM, showing that the proposed MDP + GA + RF algorithm achieved the highest accuracy of 99.00% in the Live migration of VM placement scenario, outperforming other algorithms such as KNN, DT, logistic regression, and SVM. Overall, the results of the tests conducted in the study prove that the proposed hybrid machine learning algorithm is superior to past research, achieving high accuracy and faster training times in the LVM process placement scenario, , can be seen in table 8 as follows:

Table 8. Results Algorithm ML Live Migration VM

Authors	Algorithm of Machine Learning	Allocation Scenario Test	Result Test Accuracy (%)
This paper follows	MDP + GA + RF	Live migration of VM placement	99.00

 

Reviewer#2, Concern # 5:

It is not mentioned how much it takes to train the hybrid ML (the proposed method).

Author response:  Thank you for the valuable suggestion.

Author action: We have revised the paragraph on page 7 (Section III):

3. Methods

In this section, we outline the objectives of our research and then define the problem, can be seen in Figure 3 as follows:

 

Figure 3. Flowchart of Methodology Research

Figure 3 is a flowchart that outlines the research process for the study on live virtual machine (LVM) migration using a hybrid machine learning approach. The process is divided into five main stages: introduction, literature review, data collection, model development, conclusion, and future work. The introduction stage defines the problem and motivation for the study, provides background on LVM and machine learning, and states the research objectives and questions. The literature review stage summarizes previous LVM and machine learning studies, identifies gaps and limitations in current research, and formulates hypotheses for the study. The data collection stage describes the data sources and collection methods, explains data preprocessing and cleaning, and presents the final dataset used in the study. The model development stage introduces the evaluation metrics used to assess the model's performance, compares the hybrid machine learning approach results with previous studies, and discusses the model's limitations and potential improvements. Finally, the Conclusion and Future Work stage summarizes the study's main findings and contributions, discusses the study's implications for data center performance and management, and suggests directions for future research on LVM and machine learning in data centers. Overall, Figure 3 provides a clear and concise visual representation of the research process for the study, highlighting the key stages and their corresponding activities.

Reviewer#2, Concern # 5:

For some of the references is not mentioned or not so clear the year of publication (e.g., 10, 11, 16 etc.)

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised references in manuscript article relevant research:

[10]      V. Thomas, S. K, and S. Ashok, "Fuzzy Controller-Based Self-Adaptive Virtual Synchronous Machine for Microgrid Application," IEEE Transactions on Energy Conversion, vol. 36, no. 3, pp. 2427-2437, 2021, doi: 10.1109/TEC.2021.3057487.

[11]      N. B. Kadu and P. P. Jadhav, "High performance computing with inter-virtual-machine communication," Journal of Cyber Security Technology, pp. 1-13, doi: 10.1080/23742917.2023.2256432.

[12]      R. M. Haris, K. M. Khan, and A. Nhlabatsi, "Live migration of virtual machine memory content in networked systems," Computer Networks, vol. 209, p. 108898, 2022/05/22/ 2022, doi: https://doi.org/10.1016/j.comnet.2022.108898.

[13]      Z. Ding, Y.-C. Tian, Y.-G. Wang, W. Zhang, and Z.-G. Yu, "Progressive-fidelity computation of the genetic algorithm for energy-efficient virtual machine placement in cloud data centers," Applied Soft Computing, vol. 146, p. 110681, 2023/10/01/ 2023, doi: https://doi.org/10.1016/j.asoc.2023.110681.

[14]      A. Satpathy, S. K. Addya, A. K. Turuk, B. Majhi, and G. Sahoo, "Crow search based virtual machine placement strategy in cloud data centers with live migration," Computers & Electrical Engineering, vol. 69, pp. 334-350, 2018/07/01/ 2018, doi: https://doi.org/10.1016/j.compeleceng.2017.12.032.

[15]      B. V. Srinivas, I. Mandal, and S. Keshavarao, "Virtual Machine Migration-Based Intrusion Detection System in Cloud Environment Using Deep Recurrent Neural Network," Cybernetics and Systems, vol. 55, no. 2, pp. 450-470, 2024/02/17 2024, doi: 10.1080/01969722.2022.2122008.

[16]      Y. Kumar, S. Kaul, and Y.-C. Hu, "Machine learning for energy-resource allocation, workflow scheduling and live migration in cloud computing: State-of-the-art survey," Sustainable Computing: Informatics and Systems, vol. 36, p. 100780, 2022/12/01/ 2022, doi: https://doi.org/10.1016/j.suscom.2022.100780.

 

 

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The article presented a hybrid machine learning model to estimate the live migration of virtual machine placement.

The research is interesting and the results are quite impressive when compared with the literature.

I think the authors should add the execution time of the hybrid approach to be possible to compare with the other strategies.

Moreover, It would be interesting if the authors executed their algorithm with a larger case study to see if the solution is scalable.

In section Results the authors did not explain the case study used to evaluate 5 ML models. What was the number of hosts??? Machines??? What is the impact of the increase in the number VMs or hosts??

Table 5 did not present the execution times of the hybrid model. I think it is important to know the feasibility of the approach, specially for larger case studies.

In table 8, if possible, it would be interesting to know the execution time of each approach and the size of the case study.

Line 70 -> learning machines -> machine learning

Line 103-107 -> Instead of “stage” I suggest use “Section”

Line 108-109 -> Write a paragraph between lines 108-109 explaining what will be presented in Section 2

Line 149 -> What is “0b0bbcxvbcxv xxxxv” ????

Line 323 -> What is “execution show”??? Performance evaluation??? Execution model???

Line 363 -> Figure 5 shows exactly the same information presented in table 4. Is it really necessary??? 

Line 366 -> faster is better than speediest

Line 367-> What is “show”??? Model???

 

Comments on the Quality of English Language

The english could be improved

Author Response

Reviewer#3, Concern # 1:

The article presented a hybrid machine learning model to estimate the live migration of virtual machine placement.

The research is interesting and the results are quite impressive when compared with the literature.

Author response:  Thank you for the support.

Reviewer#3, Concern # 2:

I think the authors should add the execution time of the hybrid approach to be possible to compare with the other strategies.

Author response:  Thank you for the valuable suggestion.

Author action: We have revised the paragraph on page 13 (Section VI):

For hybrid execution, we will do this in the next research, we here only focus on VM Placement with the Hybrid Machine Learning model.

The paper presents a hybrid machine-learning model for estimating the pre-copy live migration of virtual machines in a data center. The proposed model combines the Markov Decision Process (MDP), Genetic Algorithm (GA), and Random Forest (RF) algorithms to predict the order of virtual machine migration and identify the optimal host machine target. The hybrid model achieved 99% accuracy with faster training times compared to previous studies using K-Nearest Neighbor, Decision Tree Classification, Support Vector Machines, Logistic Regression, and Neural Networks. The authors suggest further research using a deep learning approach to solve other issues related to data center performance. The article presents a promising strategy for improving virtual machine migration in data centers. The hybrid model demonstrates high accuracy and faster training times compared to previous studies, indicating its potential to optimize virtual machine placement and reduce downtime. The authors also highlight the importance of considering data center performance and suggest further research. However, having more specific details on the performance of the individual machine-learning techniques and how they compare to the proposed hybrid model would be helpful. Additionally, it would be useful to see more information on the practical implementation and deployment of the proposed model in real-world data centers, for future research we will use ML execution time in the VM Placement strategy.

Reviewer#3, Concern # 3:

Moreover, It would be interesting if the authors executed their algorithm with a larger case study to see if the solution is scalable

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 13 (Section VI):

In future research, we will add execution time features to certain algorithms, so that the results can be compared with other algorithms and can see that the model is even better.

The paper presents a hybrid machine-learning model for estimating the pre-copy live migration of virtual machines in a data center. The proposed model combines the Markov Decision Process (MDP), Genetic Algorithm (GA), and Random Forest (RF) algorithms to predict the order of virtual machine migration and identify the optimal host machine target. The hybrid model achieved 99% accuracy with faster training times compared to previous studies using K-Nearest Neighbor, Decision Tree Classification, Support Vector Machines, Logistic Regression, and Neural Networks. The authors suggest further research using a deep learning approach to solve other issues related to data center performance. The article presents a promising strategy for improving virtual machine migration in data centers. The hybrid model demonstrates high accuracy and faster training times compared to previous studies, indicating its potential to optimize virtual machine placement and reduce downtime. The authors also highlight the importance of considering data center performance and suggest further research. However, having more specific details on the performance of the individual machine-learning techniques and how they compare to the proposed hybrid model would be helpful. Additionally, it would be useful to see more information on the practical implementation and deployment of the proposed model in real-world data centers, for future research we will use ML execution time in the VM Placement strategy.

Reviewer#3, Concern # 4:

In section Results the authors did not explain the case study used to evaluate 5 ML models. What was the number of hosts??? Machines??? What is the impact of the increase in the number VMs or hosts??

Author response:  Thank you for the valuable suggestion.

Author action: We have revised the paragraph on page 8 (Section 3.1):

Table 3 shows the results of the comparison among the different ML procedures utilized in LVM. The models conducted training and testing on an identical data set, and the duration of each phase was observed. The precision of each model during testing and training was also disclosed. It can be seen that the RF model has faster training and testing times, can be seen in table 3 as follows:

 Table 3. Machine learning technique evaluation (%)

ML Model	Accuracy Rate	F1 score	Time taken
SVM	92	92	989.72 s
KNN	91	91	30.05 s
DT	91	91	17.64 s
Logistic regression	89	89	0.41 s
Random Forest	93	93	335.57 s

 

Table 3 clarifies that this demonstrates a preparation precision of 100% and a testing precision of 93%. In this manner, RF gets to be the speediest and most productive show among the other models followed by the SVM model which had a high accuracy of 92%, albeit with a longer training time. The fastest time was made by KNN and DT, but with a lower accuracy level of 91% compared to RF and SVM. Later, the RF model would be combined with other algorithms.

Table 4 compares different machine learning algorithms for live virtual machine (VM) migration in terms of their accuracy in handling various allocation scenarios. The four algorithms compared are K-nearest neighbors (KNN), decision tree classification and KNN, support vector regression (SVR) models, linear regression (LR), and neural networks (NN). The first row of the table shows that the KNN algorithm achieved a test accuracy of 95.0% for the optimization of the pre-copy migration scenario. The second row shows that the decision tree classification and KNN algorithm achieved a test accuracy of 90.90% for the forecast of the workload of the VM migration scenario. The third row shows that the SVR model achieved a test accuracy of 94.61% for the optimization of the VM live migration scenario. Finally, the fourth row shows that the LR and NN algorithms achieved the highest test accuracy of 97.80% for the optimal selection of VM performance scenarios. Overall, the results in Table 4 suggest that the LR and NN algorithms may be the most effective machine-learning approaches for live VM migration in terms of accuracy, particularly for scenarios involving the optimal selection of VM performance. However, it is worth noting that the performance of these algorithms may vary depending on the specific characteristics of the data and the underlying infrastructure, so further testing and validation may be necessary before making a final decision on which algorithm to use in a given context.

Table 4. Comparison of Machine Learning Algorithms in Live VM Migration

Authors	Algorithm of Machine Learning	Allocation Scenario Test	Result Test Accuracy (%)
[17]	K-nearest neighbors (KNN)	Optimization of pre-copy migration	95.00
[36]	Decision tree classification and K-nearest neighbors (KNN)	Forecast of the Workload of VM Migration	90.90
[58]	SVR (support vector regression) model	Optimization of VM live migration	94.61
[59]	Linear regression (LR) and neural network (NN)	Optimal Selection of VM Performance	97.80

Reviewer#3, Concern # 5:

Table 5 did not present the execution times of the hybrid model. I think it is important to know the feasibility of the approach, specially for larger case studies.

Author response:  Thank you for the valuable suggestion.

Author action: For hybrid execution, we will do this in the next research, we here only focus on VM Placement with the Hybrid Machine Learning model.

Reviewer#3, Concern # 6:

In table 8, if possible, it would be interesting to know the execution time of each approach and the size of the case study.

Author response:  Thank you for the valuable suggestion.

Author action: For hybrid execution, we will do this in the next research, we here only focus on VM Placement with the Hybrid Machine Learning model.

 

 

 

Reviewer#3, Concern # 7:

Line 70 -> learning machines -> machine learning

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 2 (line 72):

I am revision already done. However, performing an LVM without considering the appropriate timing may result in an imbalanced HM workload. This statement aligns with Satpathy et al., who emphasized scheduling VM migration to prevent VMs from having excessive workloads [14]. Moreover, machine learning are considered to be pivotal algorithms capable of adopting learning systems to read data patterns as knowledge [15].

Reviewer#3, Concern # 8:

Line 103-107 -> Instead of “stage” I suggest use “Section”

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 3 (line 124-128):

I am revision already done. The content of this study requires a systematic analysis, which we will illustrate below. The second section highlights several related works on DP (data preparation). The third section focuses on the research method and data collection. The fourth section shows the results and discusses them. The fifth section presents the research conclusion, limitations, and recommendations for further study.

Reviewer#3, Concern # 9:

Line 108-109 -> Write a paragraph between lines 108-109 explaining what will be presented in Section 2

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 3 (line 129-130):

I am revision already done.

2. Related Works

In this section, we outline the theories related to the research we conducted as well as we outline the research opportunities we will conduct.

Reviewer#3, Concern # 10:

Line 149 -> What is “0b0bbcxvbcxv xxxxv” ????

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 4 (section 2.1):

I am revision already done. The authors discovered an intriguing gap in LVM improvement conducted by Haris et al. (2023) in ML, but we chose to enhance ML modeling with a more hybrid approach, making the LVM approach more optimal based on MDP and GA algorithms in decision-making and selection of LVM scheduling. This aligns with the 2020 research by Guo et al., arguing that the Markov Decision Process (MDP) is the type of NLP frequently used in reinforcement learning to provide the best policy [18]. However, the upcoming research aims to go further by developing hybrid machine learning to determine VM workloads with higher loads and presenting detailed information on LVM preparation. It involves selecting which VM instances should be migrated and when the optimal time for LVM execution is guided by optimal HM objectives. In this study, the MDP algorithm aligns with the research conducted by Alqarni et al. (2023), where problem-solving involves the reallocation of resources [30]. This hybrid ML can significantly have an impact on the prediction of overloaded VM issues and selection in scheduling processes in the LVM. In Table 1, we explore previous research conducted on LVM processes.

 

Reviewer#3, Concern # 11:

Line 323 -> What is “execution show”??? Performance evaluation??? Execution model???

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 8 (section 3.1):

I am revision already done. 3.1. Evaluation Model for Machine Learning

 

Reviewer#3, Concern # 12:

Line 363 -> Figure 5 shows exactly the same information presented in table 4. Is it really necessary???

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 9 (line 336):

I am revision already done, deleted figure 5.

Reviewer#3, Concern # 13:

Line 366 -> faster is better than speediest

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 9 (line 336-361):

Table 3. Machine learning technique evaluation (%)

ML Model	Accuracy Rate	F1 score	Time taken
SVM	92	92	989.72 s
KNN	91	91	30.05 s
DT	91	91	17.64 s
Logistic regression	89	89	0.41 s
Random Forest	93	93	335.57 s

 

Table 3 clarifies that this demonstrates a preparation precision of 100% and a testing precision of 93%. In this manner, RF gets to be the speediest and most productive show among the other models followed by the SVM model which had a high accuracy of 92%, albeit with a longer training time. The fastest time was made by KNN and DT, but with a lower accuracy level of 91% compared to RF and SVM. Later, the RF model would be combined with other algorithms.

Table 4 compares different machine learning algorithms for live virtual machine (VM) migration in terms of their accuracy in handling various allocation scenarios. The four algorithms compared are K-nearest neighbors (KNN), decision tree classification and KNN, support vector regression (SVR) models, linear regression (LR), and neural networks (NN). The first row of the table shows that the KNN algorithm achieved a test accuracy of 95.0% for the optimization of the pre-copy migration scenario. The second row shows that the decision tree classification and KNN algorithm achieved a test accuracy of 90.90% for the forecast of the workload of the VM migration scenario. The third row shows that the SVR model achieved a test accuracy of 94.61% for the optimization of the VM live migration scenario. Finally, the fourth row shows that the LR and NN algorithms achieved the highest test accuracy of 97.80% for the optimal selection of VM performance scenarios. Overall, the results in Table 4 suggest that the LR and NN algorithms may be the most effective machine-learning approaches for live VM migration in terms of accuracy, particularly for scenarios involving the optimal selection of VM performance. However, it is worth noting that the performance of these algorithms may vary depending on the specific characteristics of the data and the underlying infrastructure, so further testing and validation may be necessary before making a final decision on which algorithm to use in a given context.

 

Reviewer#3, Concern # 13:

Line 367-> What is “show”??? Model???

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 9 (line 344-361):

comparted model to scenario research. Table 4 compares different machine learning algorithms for live virtual machine (VM) migration in terms of their accuracy in handling various allocation scenarios. The four algorithms compared are K-nearest neighbors (KNN), decision tree classification and KNN, support vector regression (SVR) models, linear regression (LR), and neural networks (NN). The first row of the table shows that the KNN algorithm achieved a test accuracy of 95.0% for the optimization of the pre-copy migration scenario. The second row shows that the decision tree classification and KNN algorithm achieved a test accuracy of 90.90% for the forecast of the workload of the VM migration scenario. The third row shows that the SVR model achieved a test accuracy of 94.61% for the optimization of the VM live migration scenario. Finally, the fourth row shows that the LR and NN algorithms achieved the highest test accuracy of 97.80% for the optimal selection of VM performance scenarios. Overall, the results in Table 4 suggest that the LR and NN algorithms may be the most effective machine-learning approaches for live VM migration in terms of accuracy, particularly for scenarios involving the optimal selection of VM performance. However, it is worth noting that the performance of these algorithms may vary depending on the specific characteristics of the data and the underlying infrastructure, so further testing and validation may be necessary before making a final decision on which algorithm to use in a given context.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

In the paper titled “A Hybrid Machine Learning Model to Estimate the Live Migration of Virtual Machine Placement” authors propose ML based VM placement in cloud data center. The article needs major organization/presentation changes with the following points.

·       The title is confusing. I think the proposed method is for “ML to estimate workload and balance resources with live migration and VM placement”. This is not reflected in the title.

·       The abstract also talks about future work which is not advised.

·       The contributions should be listed in the introduction in form of points/bullets.

·       Several sentences in the article, such is the one from line 110-112 are poorly written or rephrased to avoid plagiarism while losing the meaning and context.

·       What does the title of section 2.1 mean? It does not make sense.

·       In section 2.3, the authors state that MDP is used with reinforcement learning. However, the author use MDP but don’t use reinforcement learning.

·       There is no need to define the equations for accuracy and F1 score.

·       The discussions section should be enhanced.

·       The research gap is not presented well in section 2. There are subsections in many directions. However, previous works estimating the VM migration with ML algorithms should be focused to extract research gap.

·       Some of the references are very old. Recent articles need to be added such as

Bashir, S., Mustafa, S., Ahmad, R. W., Shuja, J., Maqsood, T., & Alourani, A. (2023). Multi-factor nature inspired SLA-aware energy efficient resource management for cloud environments. Cluster Computing, 26(2), 1643-1658.

Author Response

Reviewer#4, Concern # 1:

In the paper titled “A Hybrid Machine Learning Model to Estimate the Live Migration of Virtual Machine Placement” authors propose ML based VM placement in cloud data center. The article needs major organization/presentation changes with the following points

Author response:  Thank you for the support.

Reviewer#4, Concern # 2:

The title is confusing. I think the proposed method is for “ML to estimate workload and balance resources with live migration and VM placement”. This is not reflected in the title

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the titled on page 1 (line 1)

I have corrected the title according to the reviewer's directions with the title:

Machine Learning to estimate Workload and Balance Resources with Live Migration and VM Placement

Reviewer#4, Concern # 3:

The abstract also talks about future work which is not advised.

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised abstract page 1 (abstract)

I have corrected the future work research in the abstract at the end of the paragraph:

Nowadays, the application of virtualization technology in the data center often puts more burdens on the host machine (HM), which will result in a decrease in VM performance. To describe the problem, we can use Live Virtual Migration (LVM), one of the purposes of which is to reduce the load on the VM. The paper presents a hybrid machine-learning model for estimating the pre-copy live migration of virtual machines in a data center. The proposed model uses a combination of Markov Decision Process (MDP), Genetic Algorithm (GA), and Random Forest (RF) algorithms to predict which virtual machines will be moved first and determine the best host machine target. The hybrid model achieved 99% accuracy with faster training times compared to previous studies using K-Nearest Neighbor, Decision Tree Classification, Support Vector Machines, Logistic Regression, and Neural Networks. The authors suggest further research using a deep learning approach (DL) to solve other issues related to data center performance. The article presents a promising strategy for improving virtual machine migration in data centers. The hybrid model demonstrates high accuracy and faster training times compared to previous studies, indicating its potential to optimize virtual machine placement and reduce downtime. The authors also highlight the importance of considering data center performance and suggest further research. Additionally, it would be useful to see more information on the practical implementation and deployment of the proposed model in real-world data centers.

Reviewer#4, Concern # 4:

The contributions should be listed in the introduction in form of points/bullets.

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised on page 2 listed the contribution research:

• The authors propose a novel hybrid machine learning model that combines Markov Decision Process (MDP), Genetic Algorithm (GA), and Random Forest (RF) algorithms to estimate the pre-copy live migration of virtual machines (VMs) in a data center.
• The proposed model can predict which VMs will be moved first and determine the best host machine target to optimize VM placement and reduce downtime.
• The authors compare the proposed model with existing state-of-the-art machine learning algorithms, such as K-Nearest Neighbor, Decision Tree Classification, Support Vector Machines, Logistic Regression, and Neural Networks. The results show that the proposed model achieves 99% accuracy with faster training times than previous studies.
• The authors suggest further research using a deep learning approach (DL) to solve other issues related to data center performance.

 

Reviewer#4, Concern # 5:

Several sentences in the article, such is the one from line 110-112 are poorly written or rephrased to avoid plagiarism while losing the meaning and context.

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 3 (line 131-135):

This Live Migration Model to maintain the performance of virtualization technology known as auto-scale is an optimization to increase or decrease the resource capacity in cloud computing based on workload demand by observing fluctuations. This allows applications to remain responsive and available during traffic spikes or load reductions [20].

Reviewer#4, Concern # 6:

What does the title of section 2.1 mean? It does not make sense.

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 3 (line 165-255)

In section 2.1 has been changed.

2.1. Previous studies of live migration technology

The authors discovered an intriguing gap in LVM improvement conducted by Haris et al. (2023) in ML, but we chose to enhance ML modeling with a more hybrid approach, making the LVM approach more optimal based on MDP and GA algorithms in decision-making and selection of LVM scheduling. This aligns with the 2020 research by Guo et al., arguing that the Markov Decision Process (MDP) is the type of NLP frequently used in reinforcement learning to provide the best policy [18]. However, the upcoming research aims to go further by developing hybrid machine learning to determine VM workloads with higher loads and presenting detailed information on LVM preparation. It involves selecting which VM instances should be migrated and when the optimal time for LVM execution is guided by optimal HM objectives. In this study, the MDP algorithm aligns with the research conducted by Alqarni et al. (2023), where problem-solving involves the reallocation of resources [30]. This hybrid ML can significantly have an impact on the prediction of overloaded VM issues and selection in scheduling processes in the LVM. In Table 1, we explore previous research conducted on LVM processes.

 

Table 1. Deploying Virtual Machines in Live Migration

Authors	Research Methods	Workload	Prediction targets	Key Findings
[12]	The pre-copy method	CPU	Memory Content Migration	There were no selection techniques for VM migration based on workloads.
[30]	Re-initialization and Decomposition-Whale Optimization Algorithm	CPU, Memory	Resource utilization	Only energy usage and migration costs; no specific virtual machine workloads were used.
[31]	Support Vector Regression	Memory, bandwidth, and CPU	Host utilization	It was limited to evaluating host usage; it was not suitable for evaluating the performance of VM migrations, although its precision was higher than that of other models.
[32]	Machine learning-based downtime optimization (MLDOM)	CPU, memory, and network utilization	Downtime	This process was conducted to reduce the downtime only applied to VM migration over the network environment.
[33]	Artificial neural network	large number of factors	Bandwidth usage and CPU Utilization	Performance prediction for pre-copy migration was not affected by increasing data center efficiency.
[34]	Genetic Algorithm	CPU, Memory, Network	CPU Utilization and Bandwidth	Fast VM placement was implemented.
[35]	Forecasting technique	Disk, Memory, and CPU	Disk, Memory, and CPU	Workload prediction was obtained to minimize VM migration.
[36]	Machine learning	CPU	CPU utilization	The k-nearest neighbor (KNN) is the approach and method in the decision of tree classification algorithms, which were compared to determine the number of VMs migrating with an accuracy level of 90.9%.

 

The Markov Decision Process (MDP) is an algorithm for decision-making processes and can represent an environment. Some problems in prediction can be formulated with MDP equations. MDP is the traditional formalization of sequential decision-making in which decisions have an impact on the next stage through future delayed rewards in addition to current benefits [37, 38]. Thus, MDP is associated with delayed rewards and the need to modify these delayed rewards [39].

In this investigation, the authors used MDP to delineate the support for learning between the specialist and the environment. Within the setting the multi-agent system, the MDP can be deputized as a five-tuple, where S is the state space, A is the action space,  a reward function,  is a transition state function, and  has a discount factor. The agent should choose the implementation according to its policy  and reward recipients. After that, the environment will change for the next argument of T. In a multi-agent environment, it could be observed as  a combination of local observations for N agents and  a combination of actions [40]. Meanwhile, the reward and transition functions change , . The discount reward is written using the equation . The authors’ aim to find the maximum reward in policy is expected to be written in an equation [41].

According to scholars, the genetic algorithm becomes the search technique that can solve estimates for optimization and search problems. This algorithm uses techniques, derivatives, mutations, and crossovers [42]. Genetic algorithms (AG) may have solutions to strike a balance in the accuracy vs. efficiency trade-off. The authors aim to solve scheduling in virtual machine placement using a genetic algorithm approach [43]. To begin with, the authors deliver an arrangement that will be communicated as a choice variable d with first-fit and best-fit algorithms. These arrangements are considered the beginning populace. Most issues with work planning are known to be NP-complete [44]. Within the selection, the solution with the most excellent wellness will be chosen, whereas the others will be disposed of [45, 46].

Within the selection to begin with a step, the choice likelihood is calculated for each chromosome. The likelihood of determination on the chromosome is calculated utilizing Equation 1.

=

(1)

The choice likelihood is calculated and after that separated into portions, the breadth of which is relative to the determination likelihood. The final organization includes making an irregular number and choosing a chromosome from the area that contains the arbitrary number [47]. In this hereditary algorithm, the crossover is utilized to develop the unused descendant of the chosen person within the determination arrangement. The procedures of different hybrids are examined within the writing audit. In studies using two-point crossover, the chromosome is divided into two pieces first [48]. The position of the cut on the chromosome is chosen randomly and is written in Equation 2 as follows:

(2)

 

Where  and  are two intersection points, and  Create a random number between one and the entire length of the chromosome. Subsequently, the two chromosomes between these cuts have changed their values, creating two new chromosomes. Cross-over details are shown in Figure 1.

 

 

Figure 1. Crossover [49]

Mutation operations introduce new features on chromosomes, namely exchange mutations, uniform mutations, and percussion mutations. In this strategy, the arrangement is arbitrarily chosen, and its quality esteem is supplanted by the esteem of another quality arrangement [50]. To affirm and prove that each arrangement has a comparable opportunity to induce unused highlights, the position and end position are selected first, and then the genes at those positions are swapped. In the next iteration, the position from the beginning is added and subtracted from the last position; the gene value is exchanged at this position, and so on. An illustration can be seen in Figure 2 as follows [51].

 

Figure 2. Mutation processes [44]

The Random Forest (RF) algorithm is a learning algorithm that is popularly used in regression and classification cases. This calculation combines a few choice trees (DTs) to create expectations, creating more exact results. The RF calculation can be utilized to evaluate the significance of highlights and offer assistance in inclusion determination [52]. However, since several decision trees (DT) must be created, the RF training step may be computationally costly. In algorithms based on DT, such as RF [53, 54], the Gini pollution record measures debasement in a set of cases based on their lesson names. The accomplishment is to concoct highlights that minimize the Gini list, which indicates greater homogeneity between names within the occasion lesson [55]. Mathematically, the Gini debasement record can be calculated with Condition 3.

	(3)
= 1 – [(P+ + (P-]

 

 

Reviewer#4, Concern # 6:

In section 2.3, the authors state that MDP is used with reinforcement learning. However, the author use MDP but don’t use reinforcement learning.

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 5 (line 183-200):

The Markov Decision Process (MDP) is an algorithm for decision-making processes and can represent an environment. Some problems in prediction can be formulated with MDP equations. MDP is the traditional formalization of sequential decision-making in which decisions have an impact on the next stage through future delayed rewards in addition to current benefits [37, 38]. Thus, MDP is associated with delayed rewards and the need to modify these delayed rewards [39].

In this investigation, the authors used MDP to delineate the support for learning between the specialist and the environment. Within the setting the multi-agent system, the MDP can be deputized as a five-tuple, where S is the state space, A is the action space,  a reward function,  is a transition state function, and  has a discount factor. The agent should choose the implementation according to its policy  and reward recipients. After that, the environment will change for the next argument of T. In a multi-agent environment, it could be observed as  a combination of local observations for N agents and  a combination of actions [40]. Meanwhile, the reward and transition functions change , . The discount reward is written using the equation . The authors’ aim to find the maximum reward in policy is expected to be written in an equation [41].

Reviewer#4, Concern # 7:

There is no need to define the equations for accuracy and F1 score

Author response:  Thank you for the valuable suggestion.

Author action:  already done, Equation for accuracy and F1 remove in manuscript.

Reviewer#4, Concern # 8:

The discussions section should be enhanced.

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 12 (line 447-466):

5. Discussion

The discussion section of the paper provides a detailed analysis of the results obtained from testing the development in LVM process placement. The authors performed several tests and algorithm comparisons before combining the developed algorithms. They found that the KNN algorithm had an accuracy of 92% and 91% with longer training times, while the DT algorithm and logistic regression had faster training times but lower accuracy values of 91% and 89%. On the other hand, the RF algorithm had a high accuracy of around 93% and faster training time compared to the SVM algorithm. Therefore, the authors decided to combine the RF algorithm with MDP and GA, resulting in a precision value of 99% and an F1 score of 98%, indicating that the hybrid machine learning developed was more optimal than other algorithms in the case of LVM. Table 8 summarizes the results of the algorithm ML Live Migration VM, showing that the proposed MDP + GA + RF algorithm achieved the highest accuracy of 99.00% in the Live migration of VM placement scenario, outperforming other algorithms such as KNN, DT, logistic regression, and SVM. Overall, the results of the tests conducted in the study prove that the proposed hybrid machine learning algorithm is superior to past research, achieving high accuracy and faster training times in the LVM process placement scenario, , can be seen in table 8 as follows:

Table 8. Results Algorithm ML Live Migration VM

Authors	Algorithm of Machine Learning	Allocation Scenario Test	Result Test Accuracy (%)
This paper follows	MDP + GA + RF	Live migration of VM placement	99.00

 

Reviewer#4, Concern # 9:

The research gap is not presented well in section 2. There are subsections in many directions. However, previous works estimating the VM migration with ML algorithms should be focused to extract research gap.

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 3 (line 165-182)

2.1. Previous studies of live migration technology

The authors discovered an intriguing gap in LVM improvement conducted by Haris et al. (2023) in ML, but we chose to enhance ML modeling with a more hybrid approach, making the LVM approach more optimal based on MDP and GA algorithms in decision-making and selection of LVM scheduling. This aligns with the 2020 research by Guo et al., arguing that the Markov Decision Process (MDP) is the type of NLP frequently used in reinforcement learning to provide the best policy [18]. However, the upcoming research aims to go further by developing hybrid machine learning to determine VM workloads with higher loads and presenting detailed information on LVM preparation. It involves selecting which VM instances should be migrated and when the optimal time for LVM execution is guided by optimal HM objectives. In this study, the MDP algorithm aligns with the research conducted by Alqarni et al. (2023), where problem-solving involves the reallocation of resources [30]. This hybrid ML can significantly have an impact on the prediction of overloaded VM issues and selection in scheduling processes in the LVM. In Table 1, we explore previous research conducted on LVM processes.

 

Table 1. Deploying Virtual Machines in Live Migration

Authors	Research Methods	Workload	Prediction targets	Key Findings
[12]	The pre-copy method	CPU	Memory Content Migration	There were no selection techniques for VM migration based on workloads.
[30]	Re-initialization and Decomposition-Whale Optimization Algorithm	CPU, Memory	Resource utilization	Only energy usage and migration costs; no specific virtual machine workloads were used.
[31]	Support Vector Regression	Memory, bandwidth, and CPU	Host utilization	It was limited to evaluating host usage; it was not suitable for evaluating the performance of VM migrations, although its precision was higher than that of other models.
[32]	Machine learning-based downtime optimization (MLDOM)	CPU, memory, and network utilization	Downtime	This process was conducted to reduce the downtime only applied to VM migration over the network environment.
[33]	Artificial neural network	large number of factors	Bandwidth usage and CPU Utilization	Performance prediction for pre-copy migration was not affected by increasing data center efficiency.
[34]	Genetic Algorithm	CPU, Memory, Network	CPU Utilization and Bandwidth	Fast VM placement was implemented.
[35]	Forecasting technique	Disk, Memory, and CPU	Disk, Memory, and CPU	Workload prediction was obtained to minimize VM migration.
[36]	Machine learning	CPU	CPU utilization	The k-nearest neighbor (KNN) is the approach and method in the decision of tree classification algorithms, which were compared to determine the number of VMs migrating with an accuracy level of 90.9%.

 

 

Reviewer#4, Concern # 10:

Some of the references are very old. Recent articles need to be added such as Bashir, S., Mustafa, S., Ahmad, R. W., Shuja, J., Maqsood, T., & Alourani, A. (2023). Multi-factor nature inspired SLA-aware energy efficient resource management for cloud environments. Cluster Computing, 26(2), 1643-1658.

Author response:  Thank you for the valuable suggestion.

Author action:  We have revised the paragraph on page 1 (line 44-46)

Virtual machine (VM) technology is one example of server virtualization. Some issues encountered on virtual machines include overloading, full availability of the of the host  machine (HM), overloading, and availability of the HM [7].

 

[7]      S. Bashir, S. Mustafa, R. W. Ahmad, J. Shuja, T. Maqsood, and A. Alourani, "Multi-factor nature inspired SLA-aware energy efficient resource management for cloud environments," Cluster Computing, vol. 26, no. 2, pp. 1643-1658, 2022, doi: 10.1007/s10586-022-03690-4.

 

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

All the previous comments have been addressed 

Reviewer 2 Report

Comments and Suggestions for Authors

 

The authors have responded and implemented all the remarks that were sent during the review process, which made things much clear and improved the paper.

Reviewer 4 Report

Comments and Suggestions for Authors

The authors have revised the article according to the comments.