Round 1

Reviewer 1 Report

 

This paper presents a study on resource provisioning in multi-layer networks. The study's contribution to the literature is significant as it proposes an innovative solution through the diversification of optical spectrum to improve resource utilization in modern networks.

 

Comments:

1. It would be useful to provide more details on the specific parameters and configurations used in the simulation to provide more details on the study's empirical aspect.
2. The significance of the study's contribution to the literature could be further emphasized.
3. The discussion on the potential limitations of the proposed solution could be included to provide a more balanced view of the study's findings.
4. The conclusion section could be strengthened by providing a more comprehensive summary of the study's main findings.

Author Response

Responses

The authors would like to thank the reviewers for their accurate and valuable comments and suggestions. All of them were carefully considered and addressed while preparing the revised manuscript. The resulting changes applied to the manuscript have been marked in its revised version in a separate diff file. They have also been mentioned in this response to be easily matched with a particular comment.

 

Below, we provide comments received from the reviewer along with our corresponding responses. 

Comment 1

Comment 

It would be useful to provide more details on the specific parameters and configurations used in the simulation to provide more details on the study's empirical aspect.

Response

Thank you very much for this comment. All necessary simulation details are presented in section 7 of the paper. First, the topology is presented and discussed, and then we present parameters for each layer and for analyzed traffic scenarios. We added some additional explanations to better illustrate the adopted assumptions. According to empirical aspects, we provided more details regarding the node architecture. 

Updated fragments of the manuscript (Section 7.4)

In the network layer, routers are able to groom traffic in the electric layer. Simultaneously, full flexibility is assumed in a sense that each router port is connected to a sliceable transponder able to groom multiple traffic streams in the optical domain [35].

35. Biswas, P.; Akhtar, M.S.; Saha, S.; Majhi, S.; Adhya, A. Q-Learning Based Energy-Efficient Network Planning in IP-over-EON. IEEE Transactions on Network and Service Management 2022.
    

Comment 2

Comment 

The significance of the study's contribution to the literature could be further emphasized.

Response

Thank you for your comments. We highlighted the significance of the study's contribution to the literature.

Updated fragments of the manuscript (Section10)

In cases of network congestions, efficient resource allocation in IPoEONs is significantly challenging compared to the fixed-grid DWDM networks due to increased flexibility in IPoEON architectures. In this paper, we…

Comment 3

Comment 

The discussion on the potential limitations of the proposed solution could be included to provide a more balanced view of the study's findings.

Response

Thank you for your comments. We agree that we didn’t present limitations of the proposed solution in the discussion. Following, we present the limitations of the proposed solutions. We consider only two scenarios of the selection of nodes to generate the dynamic traffic. Note that it covers only some part of the network (a small section of the network). Additionally, one might ask how the results would vary if a different set was selected. This is not relevant in terms of the validity of our results, however, we consider this aspect for future work. 

 

Updated fragments of the manuscript (Section 9)

There are some limitations that we consider for future works, for example, varying scenarios of the selection of nodes to generate the dynamic traffic. In this paper, we consider only two scenarios of the selection of nodes to generate dynamic traffic. It covers only some parts of the network. The results would vary if a different set of nodes would be selected. 

Comment 4

Comment 

The conclusion section could be strengthened by providing a more comprehensive summary of the study's main findings.

Response

Thank you very much for this comment. The conclusion section is quite short. It presents only the most important aspects of AHL and the most important summaries from the paper. According to your suggestion, we added an additional summary of our main findings. However, please note that the previous section (9. Discussion) presents the discussion on the obtained results and may also be perceived as a part of conclusion.

Updated fragments of the manuscript (Section 10)

The results obtained by us by conducting carefully selected simulation experiments confirm that AHL improves bandwidth blocking probability in a network. It is important for operators and end users, because the network is able to handle more traffic with satisfactory quality. It is especially important in cases when something unexpected occurs and it is necessary to send more traffic (more than usually) in a short time.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper proposes a resource provision scheme in SDN-based IP-over-EON multiple layer networks where part of optical resources are reserved for otherwise blocked congestion traffic. The idea is sound but not novel (which was applied in different context of networking in the past). The overall writing is acceptable, but many technical details need to be clarified and expanded. 

There are a few major technical issues need to be addressed as follows. 

1. According to the simulation settings (which should be discussed in the description of the scheme), the resources are half-half splitted between the visual and hidden provisions. This is questionable without analytical support, why equally splitted? What is the best ratio in general? Those all deserves careful investigations. 

2. Another very important missing detail is how the virtual link (for IP layer) is constructed? How do you know which lightpaths are needed before hand? Unless we are constructing a complete graph, there should be analysis and discussion on how the resource allocation and traffic pattern impact the virtual link construction. 

3. The authors shared the data and simulation. However, the current experiments are too limited. Scheme of this nature needs extensive tests on different network resource availablity, optimization goals, parameter tuning to have a convincing conclusion. 

4. Citations to the literature work should be complete or selective. In the later case, the authors should refer to the most representative or early work on respective topics. 

 

Author Response

The authors would like to thank the reviewers for their accurate and valuable comments and suggestions. All of them were carefully considered and addressed while preparing the revised manuscript. The resulting changes applied to the manuscript have been marked in its revised version in a separate diff file. They have also been mentioned in this response to be easily matched with a particular comment.

Below, we provide comments received from the reviewer along with our corresponding responses. 

Comment 1

Comment 

According to the simulation settings (which should be discussed in the description of the scheme), the resources are half-half splitted between the visual and hidden provisions. This is questionable without analytical support, why equally splitted? What is the best ratio in general? Those all deserves careful investigations. 

Response

Thank you for your valuable comment. We improved our manuscript and added explanations. Half of the optical resources is hidden, since such diversification of resources achieved the best results in our previous investigations presented in [10] and [11].  

10. Biernacka, E.; Boryło, P.; Wójcik, R.; Domżał, J. Elastic optical bypasses for traffic bursts. Computer Communications 2019, 146, 95–102. 
11. Biernacka, E. Elastic Optical Bypasses in Multi-Layer Networks. PhD thesis, 2021.

Updated fragments of the manuscript (Section 7.4)

The selection of 160 slices as hidden slices is based on the results of numerous simulations presented in [10] and [11].

10. Biernacka, E.; Boryło, P.; Wójcik, R.; Domżał, J. Elastic optical bypasses for traffic bursts. Computer Communications 2019, 146, 95–102. 
11. Biernacka, E. Elastic Optical Bypasses in Multi-Layer Networks. PhD thesis, 2021.

 

Comment 2

Comment 

Another very important missing detail is how the virtual link (for IP layer) is constructed? How do you know which lightpaths are needed before hand? Unless we are constructing a complete graph, there should be analysis and discussion on how the resource allocation and traffic pattern impact the virtual link construction. 

Response

Thank you for your comment. We added an explanation of the virtual topology. We assumed that each directed virtual link was created by the lightpath that occupied resources visible for IP. The topology of the virtual layer was the same as the physical topology. The virtual topology was static and didn't change during simulations. Background traffic is necessary to utilize resources of virtual links in varying degrees, particularly to create congestions. However, it is not taken under consideration during probability calculation and bandwidth blocking probability assessment. It represents all normal types of activity in networks, not relevant to our mechanisms. In our experiments, background traffic is sent uniformly between each pair of nodes. Background traffic is always sent over the IP layer. In other words, in the case of non-bypass, background traffic is sent through virtual links associated with lightpaths that allocate 320 slices. In cases of AHLs, background traffic is sent through virtual links associated with lightpaths that allocate 160 slices available for the virtual layer. This traffic is sent through network links according to the applied routing protocol, e.g., OSPF. Even if the capacity of links is equal, these links are occupied differently depending on the location in topology and the number of supported flows. The amount of traffic is selected in such a way as to ensure that a link is close to congestion. We believe that background traffic assumed in our experiments allowed us to analyze AHL in a near-real environment. If we do not assume background traffic then the traffic load generated by dynamic requests should be on a higher level.

Updated fragments of the manuscript (Section 7.4) 

The topology of the virtual layer was the same as the physical topology.  The virtual topology was static and didn't change during simulations. It was known in advance, which lightpaths were needed.

Comment 3

Comment 

The authors shared the data and simulation. However, the current experiments are too limited. Scheme of this nature needs extensive tests on different network resource availablity, optimization goals, parameter tuning to have a convincing conclusion. 

Response

We do not present results of resource availability since the utilization of resources and the number of O/E/O conversions can be found in our previous works  [10] and [11]. This paper extends our results from previous papers, presenting new aspects of performance analysis of Automatic Hidden Ligthpaths in Multi-layer Networks.  In the future, we will focus on optimization goals, for example, optimization of virtual topology.

10. Biernacka, E.; Boryło, P.; Wójcik, R.; Domżał, J. Elastic optical bypasses for traffic bursts. Computer Communications 2019, 146, 95–102.
11. Biernacka, E. Elastic Optical Bypasses in Multi-Layer Networks. PhD thesis, 2021.

Comment 4

Comment 

Citations to the literature work should be complete or selective. In the later case, the authors should refer to the most representative or early work on respective topics.

Response

Thank you for your comment. To the best of our knowledge, there are only a few works that are directly focused on dealing with traffic utilizing hidden resources in IP-over-EONs. According to your suggestion, we added referred papers describing diversification of resources, especially in IPoEON. We believe that this is the best choice to meet our aims. 

Updated fragments of the manuscript (Section 2)

To the best of our knowledge there are only a few works that are directly focused on dealing with traffic utilizing hidden resources in multi-layer networks

In [10], [11], [12] and [13] diversification of EON resources is explored to offload traffic fluctuations. Papers [10] and [11] focus on a path selection policy to offload traffic from congested links…

In [13] the authors present a solution for high and low priority traffic. When congestions occur, resources for AHLs are selected based on traffic priority. High priority traffic always gets the shortest lightpaths in terms of physical distance, which minimizes delays in networks. 

10. Biernacka, E.; Boryło, P.; Wójcik, R.; Domżał, J. Elastic optical bypasses for traffic bursts. Computer Communications 2019, 146, 95–102. https://doi.org/10.1016/j.comcom.2019.07.017. 11. 11. Biernacka, E. Elastic Optical Bypasses in Multi-Layer Networks. PhD thesis, 2021. 
11. Kądziołka, B.; Skała, M.; Wójcik, R.; Jurkiewicz, P.; Domżał, J. Employing FAMTAR and AHB to Achieve an Optical Resource Efficient Multilayer IP-Over-EON SDN Network. IEEE Access 2022, 10, 94089–94099. https://doi.org/10.1109/ACCESS.2022.320 4290.
12. Biernacka, E.; Boryło, P.; Jurkiewicz, P.; Wójcik, R.; Domżał J., Handling high and low priority traffic in multi-layer networks.BULLETIN OF THE POLISH ACADEMY OF SCIENCES TECHNICAL SCIENCES, submitted for publication, 2023. https://doi.org/ 10.24425/bpasts.2023.DOI. 

Author Response File: Author Response.pdf

Reviewer 3 Report

The technique of using bypass paths in networks with temporarily insufficient bandwidths to establish needed connections has been presented by the authors of this paper in a couple of other contexts, see, for example, "Survivable automatic hidden bypasses in Software-Defined Networks" (Computer Networks, vol. 133, 14 March 2018, Pages 73-89), "Elastic optical bypasses for traffic bursts" (Computer Communications, vol. 146 (2019), 95-102), "Automatic Hidden Bypasses in Software-Defined Networks" (Journal of Network and Systems Management,vol. 25 (2017), 457-480) and others. In this work reviewed, the concept mentioned above is called AHL and is studied in the context of fiber communication links, which allow operation at very high bit rates. The simulation results presented, with their good descriptions, are worthy of publication. (This reviewer recommends their publication.) These results can be used in designs of high bit rate networks under very high traffic loads.

Author Response

We would like to thank you for this positive and motivating comment on the manuscript.  Additionally, we improved our English. The resulting changes applied to the manuscript have been marked in its revised version in a separate diff file.

 

Round 2

Reviewer 2 Report

The overall paper is improved in writing, and some of my comments are addressed.  Technically, this paper is still flawed, and the contribution will be more exciting if it address the following aspects: 

1. Analytical and simulation results that shows the performance under a different resource splitting ratio.

2. The authors' response mentioned about that the traffic is static and known in advance. This assumption is not reasonable for a IP/EON network. How can you know all traffic pattern in advance? Please clarify.  

Author Response

Responses

The authors would like to thank the reviewer for his accurate and valuable comments and suggestions. All of them were carefully considered and addressed while preparing the revised manuscript. The resulting changes applied to the manuscript have been marked in its revised version in a separate diff file. They have also been mentioned in this response to be easily matched with a particular comment.

Below, we provide comments received from the reviewer along with our corresponding responses. 

Comment 1

Comment 

1. Analytical and simulation results that shows the performance under a different resource splitting ratio.

Response

The main assumption of AHLs is that the amount of hidden resources might be at most equal to the amount of resources visible for the IP layer. Assuming 320 slices as optical resources, 160 of slices is equal to half of optical resources.  In our previous works [10], [11], we proved that a less amount of hidden resources (for example 80 of hidden slices) provides higher level of BBP compared to 160 of hidden slices (available for AHLs). We decided that it is not necessary to present and repeat these results. In this paper, we mainly focus on the number of candidate paths for AHLs rather than the impact of amount of hidden resources on efficiency of AHLs. Below we present some conclusions from our previous works [10], [11]. 

In [10], [11], we present the results obtained for NSF15 and UBN24 for Vdeg and Vsp. The results show the BBP reduction gain that is defined as a difference between BBP obtained for reference scenario AHL(10) approach normalized by the BBP in the reference approach. Gain is expressed in percentage and describes the potential improvement in terms of BBP reduction by applying AHLs. Tables 1, 2, 3 and 4 summarize reduction gain for AHLs  in the NSF15 and UBN24 networks, respectively. The gain is presented in function of the traffic load and the number of hidden slices used for AHLs. In the tables, the result of 100% achieved by AHL is shown in bold. The gain equal to 100% under the given traffic load means that the request blocking event is not observed for the AHLs, whereas request blocking events appear for the reference approach. Based on the results obtained for the UBN24 and NSF15 networks, we can conclude that the reduction of BBP increases with an increasing number of hidden resources under a given traffic load. Hidden resources may be effectively utilized to set AHLs. As a result, AHLs for case 160 of hidden slices provide the best reduction of BBP as more spectrum is reserved for it. 

Based on these results we assumed that 160 of hidden slices are the best choice for assessments of AHLs also in the Euro28 topology analyzed in this paper. We did not repeat the whole experiments to confirm it, but the single, trial simulations confirmed that for 160 slices, the gain was equal to 100% under the given traffic load, like in previous cases (for other topologies).

Table 1			number of hidden slices
	Traffic	load	40	80	120	160
	40	Erlangs	80.8%	93.6%	95.7%	100%
NSF15 for Vdeg	45	Erlangs	79.0%	92.7%	93.7%	95.5%
	50	Erlangs	69.6%	87.9%	90.3%	92.4%
	55	Erlangs	65.0%	86.0%	86.6%	89.8%
	60	Erlangs	52.5%	76.6%	76.8%	83.1%
	65	Erlangs	46.9%	68.0%	68.7%	76.9%
	70	Erlangs	39.9%	58.0%	60.0%	69.7%
	75	Erlangs	33.3%	51.0%	51.2%	62.0%
	80	Erlangs	28.4%	42.9%	44.8%	54.1%
	85	Erlangs	24.5%	37.8%	40.0%	48.7%
Table 2			number of hidden slices
	Traffic	load	40	80	120	160
	8	Erlangs	72.3%	83.7%	83.8%	100%
	10	Erlangs	31.9%	76.7%	80.9%	90.2%
	12	Erlangs	29.4%	71.5%	74.9%	83.8%
UBN24 for Vdeg	14	Erlangs	21.7%	59.4%	73.9%	80.9%
	16	Erlangs	20.9%	61.8%	68.2%	75.0%
	18	Erlangs	20.7%	52.4%	67.7%	77.2%
	20	Erlangs	18.8%	49.0%	63.6%	73.7%
	22	Erlangs	17.7%	45.2%	59.6%	68.2%
	24	Erlangs	15.6%	40.3%	55.6%	65.9%
	26	Erlangs	12.0%	34.0%	48.7%	62.2%
Table 3			number of hidden slices
	Traffic	load	40	80	120	160
	40	Erlangs	95.0%	100%	100%	100%
NSF15 for Vsp	45	Erlangs	92.6%	98.9%	100%	100%
	50	Erlangs	88.7%	98.4%	99.5%	100%
	55	Erlangs	85.0%	97.3%	99.0%	99.5%
	60	Erlangs	80.1%	95.9%	98.1%	98.8%
	65	Erlangs	74.2%	93.9%	97.0%	97.8%
	70	Erlangs	67.5%	91.3%	95.5%	96.4%
	75	Erlangs	60.3%	87.3%	93.2%	94.2%
	80	Erlangs	54.4%	83.1%	89.9%	91.4%
	85	Erlangs	47.9%	77.7%	86.4%	87.9%
Table 4			number of hidden slices
	Traffic	load	40	80	120	160
	8	Erlangs	84.9%	100%	100%	100%
	10	Erlangs	78.0%	93.0%	96.2%	100%
UBN24 for Vsp	12	Erlangs	68.7%	89.3%	94.1%	95.1%
	14	Erlangs	59.9%	84.6%	91.6%	91.8%
	16	Erlangs	49.3%	77.8%	86.9%	88.5%
	18	Erlangs	41.2%	68.7%	80.9%	84.0%
	20	Erlangs	33.3%	61.0%	74.0%	77.9%
	22	Erlangs	25.8%	52.3%	65.7%	70.9%
	24	Erlangs	22.4%	45.8%	59.3%	64.9%
	26	Erlangs	17.8%	39.0%	52.4%	58.7%

10. Biernacka, E.; Boryło, P.; Wójcik, R.; Domżał, J. Elastic optical bypasses for traffic bursts. Computer Communications 2019, 146, 95–102. 
11. Biernacka, E. Elastic Optical Bypasses in Multi-Layer Networks. PhD thesis, 2021.

 

2. The authors' response mentioned about that the traffic is static and known in advance. This assumption is not reasonable for a IP/EON network. How can you know all traffic pattern in advance? Please clarify. 

Response

Thank you for the comments. We assumed two types of traffic. The first traffic type was long-lived background traffic between each pair of nodes with bit rate equal to 50 Gbit/s. Background traffic was always served in the virtual layer utilizing OSPF. The amount of traffic was selected in such a way as to ensure that a link was close to congestion.  For example, when 160 of slices were used to establish virtual links, the most loaded link in IP utilized approximately 67% of link capacity in the EURO 28 network. This allowed us to analyze AHLs in a near-real environment. The second traffic type was dynamic requests (traffic bursts) generated by nodes designated in the network according to the topology-based methods. The efficiency of the proposed AHLs was validated through numerous simulations performed for the EURO28 network under scenarios with traffic bursts generated between selected nodes. It was reasonable to investigate such scenarios since popular services are based on communication with special servers that are mostly located in selected nodes. The procedure of node selection was based on information about physical network topology. 

We highlighted that we assumed two types of traffic and modified Section 7.5.

Updated fragments of the manuscript (Section 7.5)

…. In other words, two types of traffic were assumed. The first traffic type was long-lived background traffic between each pair of nodes with bitrate equal to 50 Gbit/s….

…The amount of traffic is selected in such a way as to ensure that a link is close to congestion.  For example, when 160 of slices were used to establish virtual links, the most loaded link in IP utilized approximately 67% of link capacity in the EURO 28 network. This allowed us to analyze AHLs in a near-real environment. The second traffic type was dynamic requests (traffic bursts) generated by nodes designated in the network according to the topology-based methods [34]. The efficiency of the proposed AHLs was validated throughout numerous simulations performed for the EURO28 network under scenarios with traffic bursts generated between selected nodes. It was reasonable to investigate such scenarios since popular services are based on communication with special servers that are mostly located in selected nodes.…

[34] Goścień, R.; Walkowiak, K. Comparison of different data center location policies in survivable elastic optical networks. In Proceedings of the 2015 7th International Workshop on Reliable Networks Design and Modeling (RNDM), 2015, pp. 48–55.

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

My comments are addressed.