SDN-Based Routing Framework for Elephant and Mice Flows Using Unsupervised Machine Learning

Round 1

Reviewer 1 Report

1. There are many works conducted the similar methodology that adopted the clustering and PCA technology to differentiate the flow type used in this paper

 2. The structure of this paper needs improvement, especially in Section 3, Proposed Methodology.  

2.1 The main blocks are not clear

2.2 The following implementation part should follow the order of the proposed flow. e.g., Flow Identification part (Algo. 4) should be put ahead of the Select best path part (Algo 1, 2,3). It is suggested to use a subsection to describe each block.  

2.3 The implementation details discussed in section 4.1 can be moved to Section 3. 

3. The authors need to provide more proof of the efficiency of the proposed developed Dijkstra algorithm which combines two existing types of Dijkstra algorithm.  

4. Some references error found in the draft: Error! Reference source not found

5. The diagram resolution is low, please replace it with high-resolution pictures. 

6.  There is no need to include the Implementation/Testbed details in Section 3.1, the second graph on the Emulator, test topologies. 

7. Please provide the reference for Appendix A, Table of Features. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

1. More details for the used data set would be useful.

2. Please elaborate more on the selection of the K-means algorithm.

3. The quality of figures 7 and 8 is low and should be improved.

4. There are some compilation errors in the file. They should be corrected.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

This research introduced ML approaches for optimizing network performance and providing high Quality of Service (QoS) by prioritizing mice flows or scheduling re-routing elephant flows.


Comments:

    1. In the introduction, The authors should check this mistake, Reference numbering: 
 “One of the main causes of these problems is the unfair use of the network resources by specific flows [1], [2]. “

“ like the separation between forward-ing and control plane, control centralization, and its capability of network behavior pro-grammability [4][5]. “


    2. There is a weak background about multipath routing in this manuscript, therefore the following article deeply investigated the same purpose, the authors must consider this paper and conclude those mechanisms are considered.  DOI:
https://doi.org/10.21928/uhdjst.v4n2y2020.pp107-116  

    3. This sentence is not adequately presented “In the DCN, most of the network traffic is mice flows, which contain only several packets, while elephant flows, which represent a few of the traffic, comprise more than 80% of the total load. “  which one has a bigger packet size,  Elephant or Mice ?

4- A table should be prepared in Related work to present the mechanisms used by others authors to solve the issues and the results impact on the reduction of bandwidth consumption, and resources usages.

5- why it’s in bold “Error! Reference source not found. and Error! Reference source not found.. “  therefore the description is missing.



6- This sentence “Otherwise, it was an elephant flow and the best N paths will be chosen based on the 1/bandwidth as shown in section 4.1.6. “  why in this section the authors talked about section 4.1.6, currently it’s written in section 3?  it’s not correct.  it should be revised.

7- too many algorithms are shown in his manuscript, some of them didn’t represent novelty. The authors should fix them or use diagrams to present the contribution.

8- number of switches used for test experiments is not enough, what about 100 switches? is the implemented tool enabled to take this scenario?

9- SDN Ryu controller is not defined well nor cited inside the manuscript.

10 - why the results brutally have been changed in Figure 10. Comparison between Proposed Method and RYU Controller With Respect to Throughput for Elephant Flow , Number of flows =16? I expected that the result at that point is not adequate.

11 - Figure 11. Comparison between the Proposed Method and RYU Controller with respect To Throughput for Mice Flow, The authors should change the Ryu controller line result to the dashed line.

12 why the test result of Experiment 1 is better than Experiment 2 ? therefore the expectation of Experiment 2 should give a better result.

    13. this sentence should enhance “we created a novel SDN routing framework based on the concept of flow identification.         

 14. I can not catch the novelty of the paper in the conclusion or in the abstract, the authors should enhance the scientific quality of the paper.
   
    15.  All the short forms(abbreviations) are not defined.   

16.  The manuscript lacks a detailed description of the methodology section and a workflow diagram.

17.  The results section is fragile to support the novelty of the manuscript.

18. Manuscript lacks in comparison of the results obtained with state of art methods or models. the comparisons are not enough exclusively with Ryu. 

19. The quality of the figures can be improved. By using adobe illustrator and creating png images.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The comments have been addressed. One more comment is that the number in Fig. 8 is too small. 

Reviewer 3 Report

The authors have made the modifications correctly.