kumoh national institute of technology
Networked Systems Lab.

Review Comment

NSL > Education> Review Comment
Long - IEEE transactions of Industrial Informatics
By :
Date : 2017-07-13
Views : 4

Reviewer: 1

Comments to the Author
The article proposes a routing scheme that aim to enhance the energy consumption and end-to-end delay for large-scale industrial IoT systems based on 802.15.4a. The article is nicely written and do not contain many typos. However, this reviewer has some concerns about the article that needs to be clarified before it can be considered for acceptance into TII.

1) The title say Energy-aware real-time routing.... but this reviewer wonder where the real-time come into play in this paper and it seems it is not included in the optimization formulation? Moreover, the authors use the metric average end-to-end delay in their work. If you speaking in terms of real-time and industrial control application the metric average end-to-end delay is useless - it should be the worst-case delay! Average end-to-end delay is used in best-effort systems.

2) Regarding the simulation set-up it would be interesting to know how often packets are sent from each device since that has a huge impact on the simulations. Furthermore, has the authors used any kind of channel model and if so, what kind of model is used? This reviewer is also questioning the scenario used in the simulations. Is it actually realistic to assume 50 gateways? Most IWSNs deployed today are centralized, meaning one gateway serving X sensors/actuators.

3) I think the article would deserve from reducing the number of references and put focus on the routing scheme, results and maybe try to make a real implementation. However, this reviewer think it would be interesting for the authors too maybe compare your performance to the optimization formulation presented in the TII articles End-to-end reliability-aware scheduling for WSNs and in QoS-aware cross-layer configuration for IWSN.

Reviewer: 2

Comments to the Author
In this article, the authors propose a routing scheme that enhances energy consumption and end-to-end delay for large-scale networks.

Even though the begin their motivation for selecting IEEE 802.15.4 with M. Palattellas article, however they do not convince me why they chose IEEE 802.15.4a instead of IEEE 802.15.4e (now IEEE 802.15.4-2016), and more specifically, i.e., IEEE 802.15.4-2016 TSCH mode.

I have the impression that the authors misunderstood TDMA with TSCH, we employed a hierarchical framework consisting of three types of elements: I/O devices, routers, and gateways while an IEEE 802.15.4a carrier sense multiple access/collision avoidance (CSMA/CA) scheme is used as the MAC protocol. Instead of time division multiple access (TDMA), the IEEE 802.15.4 CSMA/CA protocol could support for satisfying either the power savings or the latency requirement in the IIoT system., see at the end of the page 1, second column.

First of all, IEEE 802.15.4-TSCH is not pure TDMA! TSCH stands for Time-Synchronized Channel Hoping, thus, TDMA + Channel hopping.
Then, TSCH was proposed so that the standard to be ultra low-power, low-latency (i.e., deterministic traffic) and highly reliable (i.e., above 99.999%). Thus, claiming that the authors chose IEEE 802.15.4a to satisfy the low-power and latency operation does not stand!

Aiming for Industrial Wireless Networks, the authors could employ the whole stack (or at least some of the RFCs) that is proposed by the IETF. For instance, reinventing the wheel, RPL, the leading routing protocol for industrial networks, should the basis to extend the potential solution in order to be layer independent.

Next, Section III C, it is not clear to me why the authors chose to heavily rely their contribution on an existing algorithm for energy efficiency purposes. As a result, I do not see any innovation on Subsection III C!

Regarding the performance evaluation campaign, I am not satisfied at all for the employed tools.
Using Matlab and OPNET Modeler, while the authors target Industrial-based applications, sounds contradictory, when there are number of freely available tools to perform the validation, i.e., Contiki OS, OpenWSN, Tiny OS, RIOT, etc. Moreover, the authors could at least employ the 6TiSCH IoT stack for their evaluation.

Last but not least, the size of the article is very short! To be more precise, the authors finished they contribution at 6 and half pages, while they dedicated more than one page for references, isnt it that much for 8-pages article ??? I would expect something between 18-22 references.

Finally, the paper have set of typos here and there, thus, it will need definitely a good proof-reading.

Reviewer: 3

Comments to the Author
The authors consider an hierarchical large-scale network. In the first tier there are devices which form clusters. In the second tier there are router that operate as cluster heads for the end devices and route traffic to the third tier, i.e. gateways. In this context, the authors propose an hierarchical routing algorithm that aims to primarily minimise energy consumption, and secondarily end-to-end delay. The authors evaluate their proposed algorithm in OPNET simulations, where the algorithm outperforms 3 other routing schemes in all domains.

The key contributions of the paper can identified in two parts of the routing scheme: the cluster head selection and the routing scheme between the router nodes. Both schemes utilise two sources of information to form the paths: the geometrical distance between nodes and the residual energy of the nodes. Given a plethora of works for energy-aware routing, the novelty of the proposed scheme is not very clear. I would suggest the authors to enumerate the key elements that make their work unique.

The authors should also provide more intuition on how the cluster head selection algorithm works. According to the scheme, the cluster head that maximised Eq. (8) is selected as the cluster head. There are two issues with Eq. (8).
- The E_{s-res_k} seems to be irrelevant, as it does not depend on the candidate cluster head r_j. In other words it contributes equal to all candidates.
- The denominator includes the average residual energy of the members of the candidate cluster. In other words, the higher the residual energy of the members of the cluster, the more likely to be selected. This is decision needs to be further justified: it is not clear how the residual energy of the cluster members affect the energy consumption of s_k or r_j.

It must be also highlighted that algorithm uses as input the geometrical distance of two nodes. This is fairly difficult to obtain: it perhaps can be estimated through the received signal strength but this estimation would be quite erroneous in realistic environments. Why the authors do not use the expected transmission count (ETX) instead? ETX is a very standard routing metric, as it incorporates the dependence of distance on link quality, the energy consumed for retransmissions and the delay caused by retransmissions.

Lastly, the authors should further justify Eq. (1). This function calculates the energy consumed for transmission as a function of the distance. This fails to capture the maximum radiation limits, as well as the fact that practical radios offer certain finite levels of transmission power settings.

Reviewer: 4

Comments to the Author
Honestly, I think the main topic of the article does not fit into this journal. I would recommend authors to send the article to other journals more specialized in the networking topic:
* IEEE Transactions on Wireless Communications
* Elsevier Communication Networks
* Elsevier Ad Hoc Networks
* ...
Or even some journal with a more general character:
* IEEE Internet of Things Journal
* ...

Reviewer: 5

Comments to the Author
The paper presents some preliminary research on energy aware and real time routing scheme for IIOT. However, the paper is not immune from major flaws.

+ the manuscript is not well prepared, a professional proof reading is required to further improve paper quality to meet TIIs basic requirement for publication. Parts of the manuscript should be re-written, the meaning conveyed was either vague or hindering the reviewers understanding of the content itself. For instance, P1. L53 the number of previous works deeply investigating solutions for developing the IIoT has been insignificant while the authors only focused on improving the IoT technology. The num. of previous works was insignificant or insufficient? The authors state that previous works focused only on improving IoT tech, where the criticism is vague, and needs further clarification. Similar problems can be found in other parts of the paper, e.g. P2. L54, In the first stage.. . Next, in the second stage; P2. L8 sentence .. characteristic of wireless communication should end up with a Period. P2. L10 Firstly, thanks to .. Finally, and so on. A thorough proof reading is required.

+Section II introduces related works, however, quite many related works have been introduced in Section I. Partitioning of the paper is not clear.

+ Readability of Figure. 4 is low.

+ Authors should strictly follow TII manuscript template when organizing Reference part.
Ref [13] page and volume information are missing.
Ref [31] paper title info. is wrong.
Ref [12] is not convincing enough to serve as reference for TII.

A thorough check is needed, page num. and publication date should be placed in correct order.

+ Papers scientific merits should be further enhanced by including detailed comparison with other related state-of-the-art works.
Ʒ ̺
Ʒ  Alif - ETFA 2017