CASM: A Cost-Aware Switch Migration Strategy for Elastic Optical Inter-Datacenter Networks

Round 1

Reviewer 1 Report

The reviewer enjoys reading the paper. The paper proposes a cost-aware switch migration (CASM) strategy for controller load balancing and focuses on improving the load balancing performance. The paper makes contributions on multiple load metrics to measure the load of controllers and identification of controller’s laod based on the controller's average response time. The load balancing mechanism may be of interests to the readers of the optical switching data center network researchers. And the paper gives some enlightenments to readers.

In general the paper is well organized, and discusses key technologies in load balancing of data center networking. However, there are several issues to be improved:

• Please describe the extend Openflow protocol and explain the difference with respect to Openflow.
• There is a typo in the 3.2: No add Nu should be No and Nu. The index of subsection “Switch Selection” should be 3.2
• Below the Fig.3, …represents the ith…, represents jth… the character should be in Italic. Please also check other parameters.
• Please comment on the scalability of the CASM mechanism.
• As for the results, the reviewer suggests to add the definition of the traffic load. Please also add the explanation on the mechanism why CASM show better performance with respect to other solutions.

Author Response

please find the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This paper proposes a Cost-Aware switch migration strategy for data centers. Compared with other SDN and migration strategies, this paper further considers the message processing overhead of the controller, and calculates the response time to evaluate the load of each controller. In addition, switch migration is performed based on evaluation results such as message processing overhead and response time. According to the experimental results, compared with the existing research literature, the strategy proposed in this paper can improve the load performance by 43.2% and reduce the migration cost by 41.5%.This reviewer believes that this paper proposes a new switch migration strategy suitable for SDN, and has a substantial contribution to the field of network communication. Minor corrections are required as follows:
1. The graph in Figure 4 should indicate what the unit is on the Y-axis
2. It is recommended to supplement the calculation method of the load imbalance degree in the graph of Figure 4.

Author Response

Please find the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

Generally, the paper "CASM: A Cost-Aware Switch Migration Strategy for Elastic Optical Inter-Datacenter Networks (Invited)" is written and organized correctly. It concentrates on a pretty important topic for inter-DCNs. The authors propose a cost-aware switch migration (CASM) strategy to achieve a balanced distribution of controller load and improve switch migration effectiveness. The proposed solution is quite interesting. However, a few flaws should be fixed before the paper is published. The main issues are as follows:

1. Subfigures (a) and (b) in all figures are denoted in bold since (c) and (d) are in regular font. Why? The authors should normalize this.
2. In line 80, there is "to to". It should be one "to".
3. In line 45, the sentence is suddenly interrupted and continued in a new line. The authors should merge both lines (45 and 45) into one.
4. In line 98, authors are citing [12], [33], and [34]. Positions [33] and [34] should be re-ordered to become new [13] and [14] positions, respectively. Other positions should then, of course, be re-numbered.
5. In line 102, it should be "in Section 5" not "In Section 5".
6. In line 268, it should be "as shown in" not "As shown in".
7. The authors should fix a lot of free space (free/empty lines) after Algorithm 1, Algorithm 2, and Algorithm 3 body. It looks very wired now.
8. In line 318, there is used a wrong subsection number. I suppose it should be 3.2.
9. Subsection "3.3 Switch Migration" should be placed on the next page, just after the Algorithm 2 body. It looks terrible because this subsection has no content just after its name.
10. Figures 4 and 5 should not be split into two pages. All subfigures for Figure 4 should be on one page - this is the same for Figure 5.
11. In Table 2, NSFNET, OS3E, Ta2, and Interoute topologies are mentioned. However, there is no mention of them in the text. The authors note them later. I suggest adding a few more sentences in this place. There should also be an explanation of why these four particular topologies were chosen.
12. In line 37, there is a reference to EONs. However, the cited articles are not the first articles that provided the EON concept to state of the art. I suggest also citing other papers mentioning first EON's topologies and ideas.

Author Response

Comment 1: Subfigures (a) and (b) in all figures are denoted in bold since (c) and (d) are in regular font. Why? The authors should normalize this.

Response 1: Thank you for your advice. We have corrected these typos and unified the formatting of these characters in the revised paper.

Comment 2: In line 80, there is "to to". It should be one "to". 

Response 2: Thank you for your advice. We have corrected this typo in the revised paper.

Comment 3: In line 45, the sentence is suddenly interrupted and continued in a new line. The authors should merge both lines (45 and 45) into one. 

Response 3: Thank you for your advice. We have corrected this typo in the revised paper.

Comment 4: In line 98, authors are citing [12], [33], and [34]. Positions [33] and [34] should be re-ordered to become new [13] and [14] positions, respectively. Other positions should then, of course, be re-numbered. 

Response 4: Thank you for your advice. Following your suggestion, we have reordered the sequence numbers of all articles in the revised paper.

Comment 5: In line 102, it should be "in Section 5" not "In Section 5". 

Response 5: Thank you for your advice. We have corrected this typo in the revised paper.

Comment 6: In line 268, it should be "as shown in" not "As shown in". 

Response 6: Thank you for your advice. We have corrected this typo in the revised paper.

Comment 7: The authors should fix a lot of free space (free/empty lines) after Algorithm 1, Algorithm 2, and Algorithm 3 body. It looks very wired now. 

Response 7: Thank you for your advice. We have fixed empty lines after each algorithm.

Comment 8: In line 318, there is used a wrong subsection number. I suppose it should be 3.2. 

Response 8: Thank you for your advice. We have corrected this typo in the revised paper.

Comment 9: Subsection "3.3 Switch Migration" should be placed on the next page, just after the Algorithm 2 body. It looks terrible because this subsection has no content just after its name. 

Response 9: Thank you for your advice. We have corrected this typo in the revised paper.

Comment 10: Figures 4 and 5 should not be split into two pages. All subfigures for Figure 4 should be on one page - this is the same for Figure 5. 

Response 10: Thank you for your advice. We have corrected this typo in the revised paper.

Comment 11: In Table 2, NSFNET, OS3E, Ta2, and Interoute topologies are mentioned. However, there is no mention of them in the text. The authors note them later. I suggest adding a few more sentences in this place. There should also be an explanation of why these four particular topologies were chosen. 

Response 11: Thank you for your advice. We have added relevant content additions and explanations. These topologies all come from typical networking structures in the industry. For example, the NSFNET topology is a backbone network structure formed by connecting supercomputing centers in the United States. Combining these actual network topologies from the industry, the performance advantages and innovation of the proposed scheme can be verified from the perspective of practical application.

Revised content—Section 4: Performance Evaluation (Page 12, Section 4 in revised paper)

“These topologies all come from typical networking structures in the industry. For example, the NSFNET topology is a backbone network structure formed by connecting supercomputing centers in the United States.”

Comment 12: In line 37, there is a reference to EONs. However, the cited articles are not the first articles that provided the EON concept to state of the art. I suggest also citing other papers mentioning first EON's topologies and ideas. 

Response 12: Thank you for your advice. We have cited other relevant literature on EON in our revised paper. Details are as follows.

Revised content—Section 6: Reference (Page 16, Section 6 in revised paper)

5. Rival, O.; Morea A. Elastic optical networks with 25–100G format-versatile WDM transmission systems. OptoElectronics and Communications Conference, 2010, 100-101. [CrossRef]
6. Talebi S.; Alam F.; Katib I. Spectrum management techniques for elastic optical networks: A survey. Optical Switching and Networking, 2014, 13, 34-48. [CrossRef]
7. Chatterjee B.; Sarma N.; Oki E. Routing and spectrum allocation in elastic optical networks: A tutorial. IEEE Communications Surveys & Tutorials, 2015, 17, 1776-1800. [CrossRef]
8. Gong L.; Zhu Z. Virtual optical network embedding (VONE) over elastic optical networks. Journal of Lightwave Technology, 2013, 32, 450-460. [CrossRef]
9. Velasco L.; Vela A P.; Morales F. Designing, operating, and reoptimizing elastic optical networks. Journal of Lightwave Technology, 2016, 35, 513-526. [CrossRef]

Author Response File: Author Response.docx