An Advanced Job Scheduling Algorithmic Architecture to Reduce Energy Consumption and CO2 Emissions in Multi-Cloud

Round 1

Reviewer 1 Report

In this paper, a two-phase process for the scheduling architecture of could computing was presented. A clustered approach utilizing K-means and Q-learning was imposed to migrate the user from one PM to another based on QoS parameters. The proposed work also incorporated CO2 emission as a major evaluation parameter other than energy consumption. Besides, the proposed work was evaluated against another recently proposed state of art techniques based on QoS parameters. However, there are still several problems to improve, which are stated as follows.

1. In the Introduction section, the motivation and contribution can be also emphasized.

2. There are many recently published literature which should be analyzed in the Introduction, such as 10.1109‬/TASE.2021.3062979‬, 10.1109/TETCI.2022.3145706, and 10.1109‬/TNNLS.2022.3208942‬.

3. Please re-organize the literature review in Section 2. Normally, the review helps know the developing history of the proposed problem and the proposed algorithm logically. In addition, the number of references is few and could be added appropriately.

4. Figure 1 in Section 3 is not clearly illustrated and described. Please try to add some more legends and sentences to describe. Meanwhile, the use of symbols should be explained clearly to avoid unnecessary conflict.

5. Please add more expressions to Table 1 to make it more standardized and clear.

6. In Section 4, to make it clear, the description of the equations should be more detailed.

7. Tables 2-4 in Section 4, the name of the comparison experiment needs careful thinking to make it more normalized.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

1) Please include a table of notations that clearly defines each term referred to in the paper. There are numerous terms and at times hard to locate the definition of these terms.

2) Please include the equations for fixed server and dynamic energy consumption. 

3) Please re-write the paragraph starting at line 241 and ending at line 251. Your claim is that 70% of the energy is consumed when the physical machine in the cloud data center is in an idle state. You also claim when an inactive physical machine is fully utilized power is significantly saved. How? From these claims, it can be deduced that idle power dissipation is greater than dynamic power dissipation. This seems to contradict established power models. Please describe the energy model clearly so that this confusion is cleared up.

4) In equation (3) refers to Table 1. How total energy is computed again is not clear. Furthermore, equations (6) and (7) also do not help in understanding the energy model. Please refer to the following papers for more details on power models:

a) A Survey on System Level Energy Optimisation for MPSoCs in IoT and Consumer Electronics

b) Energy-Aware Scheduling of Streaming Applications on Edge-Devices in IoT-Based Healthcare

5) Table 1 lists energy consumption and CO2 Emissions as the number of users increase. Why the energy consumption for 50 users is less than 60 users? Is it due to the heterogeneity of the resources? Please explain how this table has been generated. Please re-write paragraph lines 351-363

6) Updated K-means: Please explain why these two changes are required and how they improve the existing k-means algorithm.

7) In Table 1 the two approaches are compared. Please explain how the neural network approach works. 

8) Please include a few examples that demonstrate how your approach works. In the current state, it is very difficult to understand how the proposed approach works and helps in reducing CO2 emissions and energy consumption. It is very difficult for readers to understand the proposed algorithms. 

Minor Issues:

There are numerous spelling and grammatical errors. Please fix these.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

This article presents a simulation architecture having multiple clouds to share resources and load among each other. The data centers of the cloud are equipped with physical machines that handles the users in terms of their passed instruction set.This paper presents a two-stage process for cloud computing scheduling architecture. The initial phase of assigning a user to a PM has minimal cost. A clustering approach utilizing k-means and Q-learning is adopted to migrate users from one PM to another based on quality of service parameters. The proposed work also uses carbon dioxide emissions as the main evaluation parameter in addition to energy consumption. To support resource sharing, the deployment model is a multi-cloud model.

 

Pros:

-            The proposed solution to the job scheduling problem reduces overall energy consumption and CO2 emissions compared to other advanced technologies. The proposed algorithm is 10.75% higher than Hu et al. 's architecture in terms of energy consumption, and 13% better than Cai et al.

 

Minor comments:

-            Subheading 1.2 has too much white space from the paragraph below. The format should be consistent with 1.1. In addition, the format of the two subheadings is inconsistent.The formatting of the article needs further improvement.

-            On page 9, the paper mentions that the proposed algorithms differ in terms of propagation, reward and punishment generation, and the differences can be in more detail.

-            The presentation and writing of this paper need to be further improved. There are some clerical errors in the paper that must be corrected.

-       In the abstract, “put a significant impact in” should be “put a significant impact on” and “on the base of” should be “on the basis of”.

-       On page 7, “This result in” should be “This results in”.

-       On page 15, “with least proposed cost ” should be “with the least proposed cost”.

-       On page 15, “has lot of ” should be “has a lot of ”.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

In this paper, a two-phase process for the scheduling architecture of could computing was presented. A clustered approach utilizing K-means and Q-learning was imposed to migrate the user from one PM to another based on QoS parameters. However, there are still several problems to improve, which are stated as follows.

1. The motivations of the work should be described more clearly and logically.

2. Through the existing research, analyze the existing drawbacks or unsolved problems and then highlight the improvement of this paper.

3. In Table 1, what does “total number of instructions under 1 PM ‘j’” mean? In addition, “co2” means the same thing as “CO2”, right?

4. In Section 3, “The ‘j’ symbol represents…… and location” is not clear enough. For example, please explain the meaning of “ JJ ⊆ ∀ DC ”.

5. Please add numbers to the mathematical formulas in the paper, such as those in Section 3.

6. In Section 3.1, the formatting in red is confusing. Besides, please write the formula properly.

7. In Section 4, “1. Energy Consumption…… ”, and next “3. CO2 emission……”, the label order is incorrect.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

I have no further comments and suggestions.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

In this paper, a two-phase process for the scheduling architecture of could computing was presented. A clustered approach utilizing K-means and Q-learning was imposed to migrate the user from one PM to another based on QoS parameters. However, there are still several problems to improve, which are stated as follows.

1. In Section 3, “ The job set is denoted …… may vary” is ambiguous.

2. Please add numbers to the mathematical formulas in the paper, for example, “Cloud_s ⊆ ”.

3. In Section 3, “The sum of PMs …… data center”, whether “PM (R,C,L) (i j)” is “PM (R,C,L)i j in the formula “ ”.

4. The pseudo-code representation of algorithm 1 is not clear enough.

5. Please reformat the blue section in Section 3.1 and Section 4.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf