Energy-Saving Task Scheduling Based on Hard Reliability Requirements: A Novel Approach with Low Energy Consumption and High Reliability
Round 1
Reviewer 1 Report
It can be published in this form.
Author Response
Thank you very much for your approval of our manuscript.
Reviewer 2 Report
This manuscript‘s set of references lies tightly within the realm of IEEE, and the readership showing interest in the report’s finding would be within that tight grouping. This journal, Sustainability, adopts an admirable attitude towards cross-disciplinary sciences and lists over 40 in its website Aims and Scope. The topic of this paper is broadly related to the concept of sustainability but assumes much discipline specialization. The menagerie of acronyms used without explanation demonstrates the specialization of the topic. I wonder, for example, what percentage of the general readership of the journal are familiar with DVFS, and don’t need an explanation to be convinced of its utility, and indeed its popularity amongst specialists. Indeed the new algorithm introduced and advocated by this paper, ”HDFE” is used 6 times (including 4 times in the Abstract) before being spelt out finally on Line 188.
Carefully following the logic of the explanation of relevance, methodology and experimentation, the paper is useful for what is conceded as a growing group of Artificial intelligence applications which have also gradually penetrated various industries such as intelligent medicine, autonomous driving, aerospace, and security surveillance with the powerful computing power brought by multi-core processors. The paper is well constructed and Figures 3,5,5,7, and 8 are edifying. I had to reread the sentence at Line 145-146 and suggest the authors reconsider.
Author Response
Point 1: The acronym used.
Response 1: At your suggestion, we have added the full spelling of the algorithm proposed in this paper in the abstract. For space reasons, other abbreviations are not explained, but they can be found in line 84, line 108, and in the references mentioned in the red marked part of line 111.
Point 2: I had to reread the sentence at Line 145-146 and suggest the authors reconsider.
Response 2: At your suggestion, we have reviewed and reorganized the language in 146-147 as shown in the red marked section. Thank you very much for your views, which are very important to us.
Reviewer 3 Report
The authors of this paper presented energy-saving task scheduling scheme. The topic seems relevant to the scope of journal and interesting. But, there are some issues to be addressed.
1) The references seems to be consists of old publication. It is required to update the reference with recent work to reflect the state-of-the-arts in the research area.
2) English need to be polised. For example, the reviewer finds Experiments’ results in the abstract.
3) It is required to measure energy consumption as a function of deadline. How much energy are consumed if deadline is varied?
Author Response
Point 1: The references seems to be consists of old publication. It is required to update the reference with recent work to reflect the state-of-the-arts in the research area.
Response 1: Many models, theories and experimental setups in our research are based on many classical articles in the field of scheduling algorithms, and the theories proposed in these classical articles have become the agreed research rules in the field of scheduling algorithm research, so we have to refer to these classical algorithms when explaining the relevant theories. For example, reference [23], as the most classic article in the field of heterogeneous multicore, is referred to in almost all experimental setups on heterogeneous multicore scheduling algorithms. For other references to non-classical articles, we have modified them to new references such as [13][14][15][21][22]. As shown by the yellow markers on lines 86 and 101.
Point 2: English need to be polised. For example, the reviewer finds Experiments’ results in the abstract.
Response 2: Except for the changes mentioned in response 1 and response 2, the rest of the yellow marked parts are polishing of the English language.
Point 3: It is required to measure energy consumption as a function of deadline. How much energy are consumed if deadline is varied?
Response 3: Our proposed algorithm is for embedded heterogeneous multicore systems with high soft real-time reliability requirements, and there is no requirement for task deadlines in the model assumptions, but we consider the combination of real-time, energy efficiency, and reliability as a future research work. We have added considerations for future work to our conclusion, as shown in the yellow marked portion of line 339. Thank you very much for your views, which are very important to us.
Reviewer 4 Report
This manuscript proposes an algorithm HDFE by combining DVFS technology and task replication technology to solve scheduling problem of DAG applications with energy-saving and hard reliability requirements in heterogeneous multi-core processor systems. My detailed comments are listed below.
1- Figure 2: This figure is too simple, please express it in detail.
2- Table 1: Please note that the significant digits of the parameter Lambda should be same.
3- Figure 4, Figure 5, Figure 7 & Figure 8: Please check these figures. n or N in these figures? It is described points n =4 and N = 95 in L285. And there are so many different combinations of m and n. Why were these combinations chosen in these figures?
4- The verify is too weak. More tests are needed to verify the conclusions.
Author Response
Point 1: Figure 2: This figure is too simple, please express it in detail.
Response 1: This paper takes a heterogeneous multi-core processor architecture, where the individual cores on the processor are connected together by a bus. u1,u2,u3 in the figure represent three processor cores connected by a bus. At your suggestion, we have made an explanation in line 134.
Point 2: Table 1: Please note that the significant digits of the parameter Lambda should be same.
Response 2: Changes have been made.
Point 3: Figure 4, Figure 5, Figure 7 & Figure 8: Please check these figures. n or N in these figures? It is described points n =4 and N = 95 in L285. And there are so many different combinations of m and n. Why were these combinations chosen in these figures?
Response 3: The figure shows n. In different experiments we give the formula for the number of tasks n versus the number of input points N, as shown in lines 276 and 301, respectively. If you think this will cause misunderstanding among readers, we will consider changing the letter representation. In the experiment, m represents the number of processors and n represents the number of tasks. m is chosen based on the number of cores of common multicore processors and n is calculated based on the number of randomly generated input points N. This combination makes the experiment more general.
Point 4: The verify is too weak. More tests are needed to verify the conclusions.
Response 4: FFT and GE are classic practical application cases for testing scheduling algorithms, for each application we set the number of nodes in four groups (FFT: 95, 223, 511, 1151, GE: 90, 230, 527, 1175), and for each group of nodes we set ten groups of reliability requirements (0.90-0.99), and for each group of reliability requirements we generate 100 groups of applications, the total number of applications we used for testing reached 8*10*100=8000. With your suggestion, we generate 1000 sets of applications per set of reliability requirements and the new test case is 80,000. as shown in line 277 and 302 in the green marked section. Thank you very much for your views, it is very important for us.
Round 2
Reviewer 4 Report
The paper has been carefully revised. My suggestion is to accept it.
