CluMP: Clustered Markov Chain for Storage I/O Prefetch

Round 1

Reviewer 1 Report

The paper looks fine.

What is missing is the mathematics.

Especially given that the subject is genuinely (also) mathematical, and readers of the journals have or should have some closeness to math.

So please add some characteristic and meaningful formulas - novelty always appreciated.

Please use excellent formula editing.

Mind full and correct use of punctuation (also behind formulas), of italics, of blanks, of upper vs lower case, etc.

Please work with 3 experts at hand on formal and content issues.

Please find and study/compare with works (older and/or newer, articles and chapters, books, etc.) on modelling, inverse problems, image or signal processing, and addressing hybrid systems (over time), networks and hybrid systems by Erik Kropat, Ayse Ozmen, Fatma Yerlikaya Ozkurt, Betul Kalayci, Alper Cevik, Semih Kuter, Pakize Taylan, Onder N. Onak, Sureyya (Ozogur) Akyuz, N. Azevedo and Emel Savku.  If it could be done, it can become a great paper.

 

-

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper titled "CluMP: Clustered Markov Chain for Storage I/O Prefetch" presents a well-motivated approach for improving I/O processing speed in I/O-intensive tasks. The authors address an important concern in the field, considering the increasing processing speed of tasks and the relative slowdown in data transfer speed between disks and memory. The proposed CluMP method, based on a clustered Markov chain, aims to accurately predict the next block to be requested within a process, thereby optimizing prefetching.

 

The following additional points should be addressed:

 

 

1, It seems the MC mainly focuses on the immediate next most likely data to be read based on the current situation. It considers short-term data characteristics rather than being able to analyze data patterns over a longer duration. This limitation may have contributed to the suboptimal results observed in the experimental evaluation. It is important to acknowledge and discuss the limitations of the proposed approach in capturing longer-term data patterns and potential implications on the overall performance.

 

2, One aspect that could be further improved is the design details of the CluMP algorithm. While the paper provides an overview of the approach, a more detailed explanation of the clustering process and the underlying Markov chain model would be beneficial. This would help readers understand the technical intricacies of CluMP and its implementation, thus reducing any ambiguity or confusion.

 

3, The paper should provide a more detailed analysis of the data management overhead costs associated with the proposed approach. While the CluMP algorithm shows promise in improving I/O processing speed, it is essential to thoroughly examine and discuss the additional costs incurred for maintaining and managing the clustered Markov chain, such as storage requirements and computational overhead. Providing a comprehensive analysis of these costs will help readers assess the practicality and feasibility of implementing CluMP in real-world scenarios.

 

4, Furthermore, the paper should consider discussing and comparing CluMP with other existing prefetching algorithms. While the limitations of the readahead approach are mentioned, it would be valuable to provide an overview of other state-of-the-art prefetching techniques and compare their performance with CluMP. This comparison would strengthen the evaluation of CluMP and establish its superiority over a broader range of existing solutions. Additionally, the authors should include a discussion on the advantages and disadvantages of CluMP compared to these alternative algorithms.

 

5, The paper should provide more detailed explanations of the symbols and notations used in the figures. For example, in Figure 1, it is unclear what the numbers on the arrows represent and what the letters in the boxes signify. In Figure 2, the meaning of "CNx" and "Px" should be directly annotated in the figure itself, rather than only mentioned in the accompanying text. Providing clear and concise annotations in the figures will improve their understandability and ensure that readers can easily interpret the information being conveyed.

 

 

6, The experimental figures need to be redrawn. The current figures are difficult to interpret, and it is unclear what information they are conveying. For example, in Figure 9 and Figure 10, it is unclear which bars represent the "default" scenario. It would be helpful to provide clear labeling and color coding to differentiate between different scenarios or configurations. Additionally, it is important to ensure that the figures accurately represent the experimental results and convey the intended message. Redrawing the figures will greatly enhance their clarity and improve the overall presentation of the experimental findings.

Minor editing of English language required

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper of “Clustered Markov chain for Storage I/O Prefetch” is an informative and insightful research topic because this technique is not only the quality of storages generated by clustered Markov chain is advancing at I/O prefetch that is quicker than its previous rate of improvement but also support the harware engineering examine the finished board since there is still a possibility that techniques may be made in the storage solutions.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The paper improved but still needs a completion.

The paper improved.

My requests were not already done.

Letters (no other characters and no full words) in (even mini) formulas (and even in the text) must be in italics.

Add all the still missing punctuation behind formulas or items (in item lists).

See my last reports and do what you forgot or not yet did as I asked you.

Please make the outlook in full as I requested.   View it as a mini project of your team.

Reflect on (uncertainty) modelling, image or signal processing, hybrid systems, networks and systems by Erik Kropat, Ayse Ozmen, Fatma Yerlikaya Ozkurt, Betul Kalayci, Alper Cevik, Semih Kuter, Pakize Taylan, Onder N. Onak, Emel Savku and S. Zeynep Alparslan (Gok).

 

-

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The readahead algorithm in Linux is olny the widely proven and used prefetching technique. You should compare your design with other state-of-the-art prefetching techniques.

Good

Author Response

Please see the attachment.

Author Response File: Author Response.pdf