Multi-Controllers Placement Optimization in SDN by the Hybrid HSA-PSO Algorithm
Round 1
Reviewer 1 Report
Deployment optimization of the multiple controllers within a network represents an important and complex problem due to the variability exhibited by the numerous existing parameters and operating conditions. In the present study, the authors propose a hybrid metaheuristic algorithm destined to reduce data transfer volume and propagation latency by improving throughput and reliability. Several remarks can be highlighted:
1. The phrase “The existing works employ many of the heuristics approaches for controller placement problems however they limit with considering any one of the constraints (i.e., not considering controller fault tolerance, and not considering switch requirements) [14].” is confusing and needs to be rephrased.
22. At line 69, correct “[…] which is needs […]” to “[…] which needs […]”.
33. At line 85, correct “[…] most works are not considered […]” to “[…] most works are not considering […]”.
44. At lines 87 and 88, correct “[…] this research is proposed […]” to “[…] this research is proposed […]”.
55. Define all the acronyms used in the text, e.g., “CPP”, “QoS “, etc., which are not defined throughout the entire manuscript.
66. At line 107, correct “In final, […]” to “Finally, […]”.
77. Parameter HiHMS is not defined in (1).
88. About the lettering of the parameters, I propose using more habitual notations, i.e., Greek or Roman sets of characters.
9. At the beginning of a phrase or sentence, the reference number is preceded by the word “Reference”, e.g., instead of “[18] introduced a novel …”, correct to “Reference [18] introduced a novel …”.
110. At line 175, the word “SALP” must be written as a common noun “salp” (in lowercase letters).
111. What do you mean by “un optimal”, at line 189?
112. It is stated that the proposed SDN topology is constructed in a graph manner. It is not clear which of the graph properties are used in the algorithm and how the graph model comes into play within the proposed approach.
113. What do you mean by “high network overhead”?
114. In (4), what quantity is represented by lowercase “c”? The quantity “????1−?2” is denoted also as “????1↔?2” (replacing the dash with a double-headed arrow).
115. Parameter “k” is first mentioned in the text at line 354, before being defined at line 357. Also, the use of the modulus notation is not justified neither for “k” or for “V”.
116. According to (11), it is assumed that the total number of controllers is 7. Shouldn’t the theoretical background consider a generic number of controllers?
117. What type of quantity is D in (12)? The distance D is defined by itself? Is it a recurrent definition? It shouldn't be like that.
118. In relationship (14) it is not clear which quantity represents “controller attraction coefficient” and which “controller absorption coefficient”. Is there a bibliographic reference for (14)? What significance has the quantity “ℵ0”?
119. The topology described in Figure 1 is not clearly explained.
220. Also, The FA algorithm is not clearly described in Figure 2.
221. The network's characteristics are not disclosed in the problem statement section.
222. It is not clear which quantity is the actual fitness function in (16). Also, it is not stated how location and distance are considered in (16).
223. What does the upper index “x” represent in (16)? Most probably it should be “n”.
224. The PSO parameters appearing in (19) are noted differently from the same parameters initially appearing in (3).
225. What is the bandwidth unit of measurement in Table 3?
226. In (19) parameters C1 and C2 are redefined as acceleration constants. Initially, the same notations were used to denote two controllers. This inadvertent use of the same notation may produce confusion.
227. In (21) the left-hand side contains the variable 't' (whose significance is not disclosed), whereas the right-hand side does not contain the variable "t".
228. How is reliability defined? How are the numerical values shown in Table 10 being computed?
Author Response
Response Report
Response to Referee No. 1 Comments:
Dear Prof.
First, we would like to express our thanks and gratitude for your great and valuable comments. Your comments help us to improve the quality and usefulness of our paper. Regarding your comments, we have been doing the following changes:
- The phrase “The existing works employ many of the heuristics approaches for controller placement problems however they limit with considering any one of the constraints (i.e., not considering controller fault tolerance, and not considering switch requirements) [14].” is confusing and needs to be rephrased.
Response: The phrase has been revised as “The existing works employ the heuristics approaches for controller placement problems; however, they are still limited in considering some of the important constraints (e.g., not considering controller fault tolerance, and/or not considering switch requirements) [14].”.
- At line 69, correct “[…] which is needs […]” to “[…] which needs […]”.
Response: We have deleted “is” at line 69.
- At line 85, correct “[…] most works are not considered […]” to “[…] most works are not considering […]”.
Response: We have corrected “considered” to “considering” at line 85.
- At lines 87 and 88, correct “[…] this research is proposed […]” to “[…] this research is proposed […]”.
Response: We have fixed this issue at lines 87,88.
- Define all the acronyms used in the text, e.g., “CPP”, “QoS “, etc., which are not defined throughout the entire manuscript.
Response: We have added acronyms of (Controller Placement Problem “CPP”, Quality of Service “QoS “) at line 99 and 647.
- At line 107, correct “In final, […]” to “Finally, […]”.
Response: We have added (Finally) at line 107.
- Parameter HiHMSis not defined in (1).
Response: We have added parameter definition (HiHMS denotes the harmony memory solution of harmonic) at equation (1).
- About the lettering of the parameters, I propose using more habitual notations, i.e., Greek or Roman sets of characters.
Response: We have changed some parameters at equations (2) and (3). However, there are too much information and are indicated. We will work with advice in the future.
- At the beginning of a phrase or sentence, the reference number is preceded by the word “Reference”, e.g., instead of “[18] introduced a novel …”, correct to “Reference [18] introduced a novel …”.
Response: We have added the word (Reference )in proper locations in the (3. Literature Survey) section at (line 173 to 265).
- At line 175, the word “SALP” must be written as a common noun “salp” (in lowercase letters).
Response: We have fixed the word “salp” at line 175.
- What do you mean by “un optimal”, at line 189?
Response: We have revised the phrase to be “…which affected its optimality in solving the controller placement problem” to fix this issue at line 189.
- It is stated that the proposed SDN topology is constructed in a graph manner. It is not clear which of the graph properties are used in the algorithm and how the graph model comes into play within the proposed approach.
Response: We have omitted this point from the work main contribution mentioned in the introduction section as we are planning to better emphasize this issue in a subsequent paper.
- What do you mean by “high network overhead”?
Response: We mean (network load increase). The phrase is fixed at line 263.
- In (4), what quantity is represented by lowercase “c”? The quantity “????1−?2” is denoted also as “????1↔?2” (replacing the dash with a double-headed arrow).
Response: We have added( (????1↔?2 ) in (4).
- Parameter “k” is first mentioned in the text at line 354, before being defined at line 357. Also, the use of the modulus notation is not justified neither for “k” or for “V”.
Response: We have added “k” = represents the count of controllers at line 354. Also, we have deleted modulus notation of “V” and “k = U” at line 357.
- According to (11), it is assumed that the total number of controllers is 7. Shouldn’t the theoretical background consider a generic number of controllers?
Response: We have added (“,… Fcn “) according to (11).
- What type of quantity is D in (12)? The distance D is defined by itself? Is it a recurrent definition? It shouldn't be like that.
Response: We have added (distance between controller, Dc) in (12) at line 382.
- In relationship (14) it is not clear which quantity represents “controller attraction coefficient” and which “controller absorption coefficient”. Is there a bibliographic reference for (14)? What significance has the quantity “ℵ0”?
- Response: We have added and clarified the amount of the controller's attraction coefficient (the relatively high light absorption coefficient of a firefly) at line 387 and (to attract the best controllers) at line 389.
- Response: Bibliographic reference to (14) such as:
Liu, Fagui, and Dexiang Zhong. "GSOS-ELM: An RFID-based indoor localization system using GSO method and semi-supervised online sequential ELM." Sensors 18.7 (2018): 1995.
Mo, Yuan-bin, Yan-zhui Ma, and Qiao-yan Zheng. "Optimal choice of parameters for firefly algorithm." 2013 Fourth International Conference on Digital Manufacturing & Automation. IEEE, 2013.
- Response: We have added and definition ( ℵ0=Best High Light Intensity for Firefly) in (14).
- The topology described in Figure 1 is not clearly explained.
Response: We have added Figure 1 represents (the overall system model of the proposed MC-SDN architecture with distribution mechanism and selection of controllers by FA algorithm).
- Also, The FA algorithm is not clearly described in Figure 2.
Response: We have added Figure 2 represents (optimal controller selection using FA depending on the best firefly with high light intensity that appear from their abdomen).
- The network's characteristics are not disclosed in the problem statement section.
Response: We have added (The proposed MC-SDN model can be considered to be applicable to generic SDN network characteristics and can be used in efficient, flexible, scalable and reliable manner) at (4. problem statement) section.
- It is not clear which quantity is the actual fitness function in (16). Also, it is not stated how location and distance are considered in (16).
- Response: We have added equation as:
where Fitness = 1/∑d=1 to B (f(h)d (16)
- Response: Also, we have added (fitness function is evaluated based on location and distance B, d parameters to provide efficient optimal solution as follows) at line 416.
- What does the upper index “x” represent in (16)? Most probably it should be “n”
Response: We have replaced the upper index “x" to the upper index (n)at (16).
- The PSO parameters appearing in (19) are noted differently from the same parameters initially appearing in (3).
- Response: We have replaced equation (3) (PSO) section by equation (19)
In addition, the velocity updating is performed by considering inertia weight to the updated formula of velocity which is expressed as follows,
Vij(R+1) = W(R)vij (R)+ A1R1+(P_Best ij – Pij(R) )+A2R2( GBest – Pij(R ))
with parameters appearing from (line167 to 169 ).
- Response: Also we should replaced equation (2) in (PSO) section by equation (20) =Pij(R+1)=Pij(R) +Vij(R+1)
with parameters appearing at (line 160 and 161 ).
- What is the bandwidth unit of measurement in Table 3?
Response: We added the unit of measurement (Mbps) for bandwidth in table 3.
- In (19) parameters C1 and C2 are redefined as acceleration constants. Initially, the same notations were used to denote two controllers. This inadvertent use of the same notation may produce confusion.
Response: We replaced parameters C1 and C2 to parameters (A1 and A2) in equation (19) and added the parameters appearing at (line 436).
- In (21) the left-hand side contains the variable 't' (whose significance is not disclosed), whereas the right-hand side does not contain the variable "t".
Response: We have deleted variable 't' at (21).
- How is reliability defined? How are the numerical values shown in Table 10 being computed?
Response: We added the definition of reliability (RL) with equation number (26) and parameters appearing keeping the sequences in section (6.3.5. Impact of Reliability) at line (601)
RL=max∑v ∈ V ∑s∈S P(v,s)
RL represents the reliability to minimize the control path between the MC-S and indicates the available probability of the control path.
Reviewer 2 Report
1. Combine Open flow together. OpenFlow is the correct abbreviation.
2. In line 44, change improves to "It improves". 3. At line 55, give a gap after the reference number. 4. There is no distinction on what kinds of SDN networks are considered: in-band or out-of-band. Consider reading the work on in-band control, queuing and failure recovery functionalities for OpenFlow. 5. There is no need to create subsections in the introduction. The subsections are small so avoid adding those. 6. In the literature survey, create a table describing all the literature surveys with a discussion on advantages and disadvantages. 7. The rest of the paper is fine.Author Response
Response to Referee No. 2 Comments:
Dear Prof.
First, we would like to express our thanks and gratitude for your great and valuable comments. Your comments help us to improve the quality and usefulness of our paper.
Regarding your comments, we have been doing the following changes:
- Combine Open flow together. OpenFlow is the correct abbreviation.
Response: We have corrected (OpenFlow) at line 40 and everywhere else.
- In line 44, change improves to "It improves".
Response: We have added ("It improves") at line 44.
- At line 55, give a gap after the reference number.
Response: We have added a gap at line 55.
- There is no distinction on what kinds of SDN networks are considered: in-band or out-of-band. Consider reading the work on in-band control, queuing and failure recovery functionalities for OpenFlow.
Response: We have considered the work on in-band control. The proposed method can be considered to be generic in this respect. However, more exploration of this point will be considered in a future work, as highlighted in the (Conclusions and Future Work) section.
- There is no need to create subsections in the introduction. The subsections are small so avoid adding those.
Response: We have deleted the subsections in the introduction.
- In the literature survey, create a table describing all the literature surveys with a discussion on advantages and disadvantages.
Response: We have added table 2 at the end of (3. literature surveys) section at line 274 to summarize the advantages and disadvantages of the literature survey.
- The rest of the paper is fine.
Response: Thanks.
Reviewer 3 Report
The paper has a good potential for being appreciated and cited, but it requires a little improvement.
The main motivation of this study must be clarified in the introduction section to facilitate future readers. And then they can find the main idea and how the given problem has been solved. Try to put your contribution in a challenge and solution manner, in which you show the problem of the existing studies and your solution which is the new contribution point. To better delineate the context and the different possible solutions, you can consider https://0-www-mdpi-com.brum.beds.ac.uk/2079-9292/10/18/2250 and https://0-link-springer-com.brum.beds.ac.uk/article/10.1007/s10723-007-9062-z.
The future scope of the methodology should be extended/highlighted.
Author Response
Response to Referee No. 3 Comments:
Dear Prof.
First, we would like to express our thanks and gratitude for your great and valuable comments. Your comments help us to improve the quality and usefulness of our paper.
Regarding your comments, we have been doing the following changes:
The main motivation of this study must be clarified in the introduction section to facilitate future readers. And then they can find the main idea and how the given problem has been solved. Try to put your contribution in a challenge and solution manner, in which you show the problem of the existing studies and your solution which is the new contribution point. To better delineate the context and the different possible solutions, you can consider https://0-www-mdpi-com.brum.beds.ac.uk/2079-9292/10/18/2250 and https://0-link-springer-com.brum.beds.ac.uk/article/10.1007/s10723-007-9062-z.
Response: We have added (Designing an efficient methodology for optimizing the multi-controllers’ placement in an adaptable manner is very important and challenging issue in such a complex environment, especially due to the various related requirements. Thus, optimizing the deployment for multiple controllers within SDN is a complex issue due to the variability exhibited by many parameters and current operating conditions.) to clarify the main motivation in the introduction section. Indeed, a revision for this section has been done after deleting the subsections.
The future scope of the methodology should be extended/highlighted.
Response: The (Conclusions and Future Work) section has been revised in response to this comment.
Round 2
Reviewer 1 Report
The paper has significantly gained clarity and quality. In general, the authors have properly addressed my comments.
Author Response
Response Report – Round 2
Response to Referee No. 1 Comments:
Dear Prof.
First, we would like to express our thanks and gratitude for your great and valuable comments. Your comments help us to improve the quality and usefulness of our paper. Regarding your comments, we have been doing the following changes:
(x) English language and style are fine/minor spell check required
The paper has significantly gained clarity and quality. In general, the authors have properly addressed my comments.
Response:
A language revision has been done to fix some minor linguistic and style issues.
