Task Allocation Methods and Optimization Techniques in Edge Computing: A Systematic Review of the Literature

Round 1

Reviewer 1 Report (Previous Reviewer 2)

This new version is a clear improvement to the previously submitted manuscript.

The authors have done a much better job in elaborating about the literature collection process, clearing up the confusion with the deep learning compression and overall documenting their work.

With that being said, there are still minor issues for improvement.

1. RQ2 mentions that the most "promising" works are selected. please define "promising". In general,  every metric must have some sort of quantitative scale and the authors should elaborate on this.

2. some of the exclusion criteria needs a deeper overview to understand what is left behind when "excluding" entries from the corpus. For example, what does the reader (in this case someone interested in task allocation) lose when the authors take out the literature not available from their university server? please follow this for the other criteria as well.

 

Author Response

Please see our response in the attached file.

Author Response File: Author Response.pdf

Reviewer 2 Report (Previous Reviewer 3)

Remowe figure 1

Figure 2 not clear and visible

Figures 3-5 should be smaller

Remowe figure 1

Figure 2 not clear and visible

Figures 3-5 should be smaller

 

Author Response

Please see our response in the attached file.

Author Response File: Author Response.pdf

Reviewer 3 Report (New Reviewer)

The subject of this paper is contemporaneous and very relevant. However, some relevant aspect must be observed, such as:

a) Abstract: until line 9 all the comments are well-known to every system. In other words, those could be applied to other environments. It is necessary to have a more specific characterization for the Edge configurations.

b) Introduction: It is NECESSARY to have a picture characterization of an EDGE environment, where FOG and CLOUD also appear. Nowadays, a large amount of people did not realize yet that the EDGE-FOG-CLOUD is the appropriate infrastructure to modern challenges in computing scenario.

The paper requires this proposed figure to guide readers, some parts of the contribution will be much (much) clear.

c) It is also suggested to insert more figures to explain some questions of the survey.

f) Finally, an especially important aspect must be observed:

- Who is the element which is going to request (and why) task allocation and others' services in the Edge Layer? 

Be aware that some parts of the text are difficult for ordinary readers.

 

 

The text requires some careful review in terms of more easy/clear information target.  Again, task allocation and work request require a clearer explanation.

It is suggested figures to help the understanding of the text, in which EDGE-FOG-CLOUD is the computing environment

Author Response

Please see our response in the attached file.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report (New Reviewer)

This version of the contribution brings a more clear scenario from the survey.

An example is figure 1, which is much more illustrative than the previous older figure 1.

The paper brings an interesting survey which could provide a basis for several other researches. 

I would be interesting a final reading from the authors to check if the target message is all right to readers.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

In this paper, the authors present a systematic literature review of task allocation methods and related optimization techniques in edge computing. The reviewer has the following concerns about the manuscript.

First, there have been numerous surveys about task allocation/scheduling in edge/fog computing published recently. It is not clear what novelty the literature review presented in this paper has.  A thorough comparison between this survey and the related surveys is needed to elaborate on the novel contributions made in this paper.

The contents of the literature review fail to answer some of the research questions defined by the paper. For example, the reviews provide little insights to answer RQ2 -- "What are the most promising approaches for optimizing task allocation in edge computing?" The authors just described each of the works selected to be included in the review without providing any comparison or taxonomy of the works; therefore, it is hard for readers to understand the advantages and limitations of the reviewed work and obtain insights about which methods are most appropriate in various edge computing scenarios. 

It is confusing to have "how deep learning compression is performed in task allocation methods in edge computing" as one of the research questions, since "deep learning compression" is just one specific technique that is employed by just a group of task allocation methods and thus is irrelevant to the other methods (even some ML-based methods).

This paper catigoriezes task allocation methods into some overlaps groups using criteria that are in different dimensions. Distributed allocation methods are identified using the control structure as the critieria (with centralized methods as the counterpart) while machine learning-based allocation methods are identified according to the employed techniques (with optimization-based methods and game theoretic methods as peer groups). This paper only reviewed distributed methods without any discussion about centralized methods. Similiarly, the paper has one section for dynamic allocation but no discussion about static allocatin as its counterpart. Overall, the categorization of the reseaarch works and the organization of the presentation in this paper are confusing.  

 

Reviewer 2 Report

The authors of this work introduce a survey paper on the research landscape of task allocation in edge computing.

A positive note for this survey is the vast literature overview that has been conducted.

The following are comments for improving the quality of the work:

-          The authors mention that there are several review and survey studies on the topic and this is quite true. Based on this, the authors must include a dedicated section on examining the purpose, methodology and findings of these in comparison to the current work in order to understand the benefit of having one more survey.

-          The authors at one point early on mention that a comparison will be performed against the examined works and this kind of makes the reader expect a quantitative comparison which is positive but then finds out that at most a quantitative comparison is all. Please make this extremely clear even in the abstract or title.

-          The survey paper lacks specific findings and a systematic way of comparing the studied papers. A taxonomy must be extremely clear and the strengths, limitations along with other examined metrics must be clear so that any reader (put yourself in the reader shoes) will read the paper with certain criteria in mind and will know what technique(s) will consider.

-          In terms of the research questions, RQ3 and RQ4 are rather strange and the authors must elaborate a bit more on these. For example, cant an IoT network be “stitched” together using 5G and on the other hand, why is DL compression so important as a high-level research question?

-          In terms of the methodology, I understand using scopus and keyword search but why not go to the premier conferences and journals and go through these? At least if scopus is used, please provide an overview of where these studies are found (i.e., ccgrid, sec, ucc, ic2e, edge, tpdps, tiot, etc). Furthermore, some keywords make littles sense and the authors should elaborate a bit more. For example, why are uavs so relevant but not other domains like bigdata or data stream processing where task allocation mechanisms are actually quite similar.

-          Finally, the authors must introduce a future directions sections for such a survey and highlight the roadmap for the landscape.

Reviewer 3 Report

The manuscript presents the review of works on task allocation methods and optimization techniques in edge computing.

The drawback of the manuscript are a follows:

1. the title is not correct – it suggests the new solutions/ideas inside – it is not true – the manuscript present the survey on the topic – the title has to be changed

2. figures 1, 3, 4, 5, 6 are too big – they have to be decreased!!!!! moreover their contribution are rather weak – I proposed to presents such data in the form of tables

3. remove figure 2 – what is its contribution?

4. content of section 4 is not adequate to it title – moreover this content has to be presented as section 2

5. where is the analysis of the survey’s results made according to RQs? I can’t find it – it is the most important part of such type of papers

6. also the list of changes should be presented as the conclusions to the made survey

7. there is no discussion of made survey

 Moderate editing of English language is required