DRL-Based Dependent Task Offloading Strategies with Multi-Server Collaboration in Multi-Access Edge Computing

Round 1

Reviewer 1 Report

The article presents a new strategy for the offloading problem in mobile edge computing of multi-user and multi-server. 

The related works are clearly described and the problem is detailed presented. I think the work presented is a very 

interesting and relevant. 

 

In the Related Work section or in other part of the article I think the authors could present a table with the related

work comparing the characteristics of the solutions. For example, the problem that each related work solve, the strategy 

used, drawbacks. The last line could present the solution proposed with its advantges in relation with other works.

 

As the authors say in the results section,  current real datasets contain information for only a very limited number of

applications. Therefore, I think the authors should mention in the article that to use their solution it is NECESSARY a 

to create/generate a DAG of the application, if it not exist already.

 

 

 

Introduction

 

The authors say that the offloading problem is NP-hard. As they did not prove it in the article a reference wuold be necessary.

 

 

In the second paragraph of the introduction the authors use the initials DRL without writing the meaning.

 

 

Some sugestions

 

Figure 1 could be centralized

 

I think the authors should write a paragraph between Setion 4 and 4.1 explaining what will be presented in the subsections.

 

In algorithm 1 I did not understand the two horizontal lines between lines 15 and 16

 

I think the authors should write  a paragraph between Setion 5 and 5.1 summing up what will be presented in the Section.

 

I think tha authors should choose the word it will use "Fig." or "Figure"

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This work focuses on the task offloading for mobile edge computing, with deadline and task dependence constraints. The authors formulate the offloading problem as a discrete optimization problem, minimizing the weighted sum of normalized makespan and energy consumption. And authors exploit Soft Actor-Critic deep reinforcement learning algorithm for solving the offloading problem. Simulated experiments are conducted for verifying the performance of their proposed algorithm. There are some minor issues should be addressed before this work's publishing.

(1) Reasons of experimental results should be discussed in detail.

(2) Resolutions of figures should be improved.

(3) The outline of Algorithm 1 should be presented using algorithm package of latex, instead of in an image.

(4) Following related works published recently are recommended to be cited to improve the related work. (updated refs 12.16)

Liu, Sige, et al. "Contextual User-Centric Task Offloading for Mobile Edge Computing in Ultra-Dense Network." IEEE Transactions on Mobile Computing (2022). Wang B, Cheng J, Cao J, Wang C, Huang W. Integer particle swarm optimization based task scheduling for device-edge-cloud cooperative computing to improve SLA satisfaction. PeerJ Computer Science, 2022, 8:e893 https://doi.org/10.7717/peerj-cs.893

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

1. The corresponding author should not use the QQ email address but the email address of your university. 

2. The abbreviation DRL of the "deep reinforcement learning" in the abstract should be given in order to echo the DRL in the title. 

3.In the latest international communications standards, the full name of MEC has now been changed to Multi-access Edge Computing. 

4. "SACDTO (Soft Actor-Critic based Dependent Task Offloading) " should be "Soft Actor-Critic based Dependent Task Offloading (SACDTO) ".

5. The first chapter lacks the title "I. Inroduction" and there are no references in the first chapter.  The author should follow each statement with appropriate references. 

6. The authors claim six major contributions, which is a bit of an exaggeration.  It is recommended to consolidate the major contributions to 3-4. 

7. Some figures must be redrawn. Many figures in this paper are not clear and have low resolutions, e.g. Fig.1, Fig. 2.

8. Variables in the body should use the same format as variables in the equation, such as italics. 

9. Algorithm 1 should use text format instead of image format.

• 10. The sources of reference are too concentrated.  It is suggested to replace some relevant papers from other publisher, such as MDPI, Springer, etc. 

Author Response

请参阅附件。

Author Response File: Author Response.docx