Adaptive Computation Offloading with Task Scheduling Minimizing Reallocation in VANETs

Round 1

Reviewer 1 Report

Highlight novelty of proposed study clearly in Abstract.

Remove  algorithm1 & algorithm 2 from related work. Only provide references.

Problem formulation is sufficient in section 4 but methodology is not discussed properly.

Write pseudo code of  algorithm 3,4 & 5 properly like algorithm 1 & 2.

Complexity analysis based comparison of proposed algorithm is required in the result section.

Technical language editing is required.

Please modify your conclusion part as per your findings.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper proposed a vehicle adaptive offloading algorithm considering the relationships between tasks in a directed acyclic graph and the topology change of VANETs for scheduling the computationally intensive tasks. The following comments are for further considerations:

There are some minor English problems (e.g., single/plural) throughout the manuscript, please check them all in the revised version.

Line 58, page 2, the authors mentioned the relationship between the tasks. Did this paper process several tasks under a batch during a time interval or several subtasks in a given task arrived the network at a time? This should be described in the text. Next example shows that there are dependencies between tasks. Thus, the authors should indicate possibly practical applications of these kinds of tasks and their significance.

Line 66, page 2, which network component will construct a DAG of the tasks? If RV is responsible for doing that, how can it achieve the global topology and/or information of neighboring vehicles? Assumed that it can, do you think that this permission could raise a lethal security problem since one vehicle can have all information about its surround vehicles? If such security problem is ignored, the mechanism is able to impose lots of overheads between vehicles for joining into the CO?

Line 71, page 2, VAO is required to install in each vehicle since any vehicle can become an RV. Hence, at a time, there might have more than one vehicle conducting CO. How do we handle this conflict?

Is it efficient when each RV (amongst several RVs) is collecting the data in real-time, processing the retrieved information, then builds DAG and assigns the tasks? After that, it also needs to control the completion of tasks by observing the locations of current SVs and reallocate if needed. Is it required “excessive computation from a single vehicle”?

There might have the case that existing vehicles cannot handle the given task due to limited capability, is the task allowed to split into several parts that can be processed at different SVs?      

In related work section, the reviewer personally thinks that section 2.1 should mention the limitations of the existing works along with their contributions. In addition, section 2.2 would survey the task scheduling in VANET considering DAG, which has been well investigated. The latest research work was from 2019, so the authors should update this section with newer papers.

In Section 5.2, Algorithm 4, the authors should explain the whole algorithm instead of stopping at line 4. Furthermore, how to identify the value of alpha blocks? Why blocks instead of actual distances? What happened if a SV is out of the communication range of the RV but the blocks are less than alpha?

In Algorithm 5, please describe how to recognize that a vehicle is “expected to leave the communication range”? Such vehicle needs to send a signal back to RV to leave the network or RV actively measures the current distance between them? It seems that Algorithm 5 is required to make a reschedule plan for the uncompleted tasks after communicating back-and-forth with SVs. How do the authors guarantee that total procedures including RV-SVs communications and schedule-reschedule plans do not violate the delay requirements of the tasks? The problem might get worse since the tasks are processed sequentially. What if the changes of locations of SVs are faster than the available-SVs update of the algorithm?  

In Section 6, HEFT [7] and Max-Min [8] are not absolutely new since they were proposed more than 10 years ago. Thus, performance results produced by comparing with them are not persuasive. The authors should compare with any newer algorithms to convince the readers.  

The authors claimed that VAO OUTPERFOMED HEFT when |V| was greater than 8, which is not indeed convincing as shown in Figure 6a since we can see that they intersected at |V|= 8 and tended to be better a bit compared to HEFT. It might need more data (e.g., |V| > 8) to prove that conclusion.   

How to calculate the probability of task reallocation in Figure 7 and which figure depicts that the schedule length of VAO was better than HEFT by 14.4%?

Lines [398-412], page 14 might not attract the readers for applying VAO whereas Figure 9 should be analyzed more and linked to Figure 6, specifically when |V| is greater than 5.  

It is supposed that the explanations of the performance results in all figures should be further described, focusing on how improvements VAO can achieve.

It would be better if the authors compare VAO with the state-of-the-art task scheduling algorithm.  

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper is well written and organized. The authors propose vehicular adaptive offloading for the dynamic vehicular environment based on the DAG scheduling for optimizing task execution. The paper has a good contribution to the body of knowledge. However, some minor improvements are required:

• A table for comparing the related approaches is required to emphasise the novelty of the proposed approach.
• Table of simulation parameters and the simulated scenarios need to be included.
• Figure 4 is unreadable and the utilized type of graphs in Figures 6 and 9 could be changed. 
• Limitations and future works need to be addressed clearly.
• References must be updated.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

All the suggestions have been incorporated by authors. 

 

Reviewer 2 Report

Thank you for the authors' responses. 

Generally, the revised version somehow meets my expectations but I feel that the authors did not answer my concerns directly, especially essential questions (e.g., comment [1-3, 5-6]). Some editorial mistakes have easily found in their responses. I advise that the authors should focus on the main points of the comments instead of going around the bush, and are insist of their texts in the previous manuscript, which do not help at all. I personally think that the authors did not answer the first 3 comments well and they should have spent more time on my comments. Your paper could have been impressive if the authors knew how to respond to the reviewers' concerns properly. Fortunately, the results would save the whole paper as they can be used for references.