ZDC: A Zone Data Compression Method for Solid State Drive Based Flash Memory
Round 1
Reviewer 1 Report
Authors answered to all my comments except the one referring to the CDF percentile considered for the latency graphs. Is it the median or another one in the tail? Moreover I suggest the authors to improve the reference list by adding works describing the design methodologies for optimal SSD performance which are important to provide a context for the work here presented.
Author Response
Point 1: Authors answered to all my comments except the one referring to the CDF percentile considered for the latency graphs. Is it the median or another one in the tail? Moreover I suggest the authors to improve the reference list by adding works describing the design methodologies for optimal SSD performance which are important to provide a context for the work here presented. 

Response 1: Thank you for your suggestion. The latency graphs are the median. It represents the time consumed by each method when different traces are entered. I have added content and references for SSD performance optimization. SSD performance improvement methods mainly include I/O performance optimization, garbage collection method optimization and address mapping table scheduling optimization. The main corrections can be found in the revised manuscript in GREEN color.
Reviewer 2 Report
The manuscript proposes a method of solid-state drive (SSD) data deduplication based on zone division. It divides the storage space of the SSD into 2 zones of hot zone and multiple cold zones according to the different erasing frequency. The address mapping table in the hot zone is loaded into the cache. The authors did simulation tests, the correctness and effectiveness of this study are verified. The results have reference value for improving the performance and reliability of SSD. The manuscript is representative enough and it is organised in proper way. The reviewer votes for accepting at current state.
Author Response
Point 1: The manuscript proposes a method of solid-state drive (SSD) data deduplication based on zone division. It divides the storage space of the SSD into 2 zones of hot zone and multiple cold zones according to the different erasing frequency. The address mapping table in the hot zone is loaded into the cache. The authors did simulation tests, the correctness and effectiveness of this study are verified. The results have reference value for improving the performance and reliability of SSD. The manuscript is representative enough and it is organised in proper way. The reviewer votes for accepting at current state. 

Response 1: Thank you for your suggestion. We have modified the content of the full text from the title to references as well as format according to sample paper published online. The major revised portions are marked in green in the content, but without being limited to. And in order to minimize language-related errors, the paper has been polished by a professional English polish service, which we hope meet with the approval.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
The paper presented by the authors proposes a system for compressing data in SSD in order to improve write amplification and performance through zoning. The work has several flaws in present form and cannot be recommended for pubblication:
- The overall system of Fig.1 is badly explained. What are the single functions implemented in CPU? Which GC algorithm is considered? Which wear-leveling? Static or dynamic? Is it zoned?
- The system cannot work without an ECC in state-of-the-art SSD. Could you please mention why this important part of hardware in the SSD controller is not described?
- Tables 1 and 2 are not easy to be understood. Data are all zeros...
- LZW algorithm is not new in SSD controllers. What is the real added value here besides the parallelization?
- What are the characteristics of the simulated SSD? The extremely high variability both in the workload and in the internal NAND Flash characteristics could make your algorithm fail. Could you elaborate on this?
- Fig.5 is referring to latency. The median one or a particular point in the tail of the latency CDF distribution?
- English grammar should be improved
Reviewer 2 Report
The paper has serious presentation issues. There are many grammatical mistakes, incomplete sentences, faulty sentence structure as follows:
Page-1
Data compression in SSDs is where the SSD controller compresses the data before writing it to flash memory Data compression can reduce the amount of real data written, which can improve the performance and reliability of SSDs.
Page-2
“At the same time, since the number of program/erasures (P/E) cycles of the blocks in the SSD is limited, the number of P/E cycles of the SLC architecture is about 100K, the number of P/E cycles of the MLC is about 10K and the number of P/E cycles of the TLC is about 1K [5].”
Page-5
“improve SSD. Performance”
Page-5
However, SSD is a kind of data storage device, and its stored data has a certain locality in time.
Figure-3
The flow chart needs to be labeled correctly. For example, the “yes” and “no” on the decision on “cold data” needs to be swapped.
In general, more information is needed on the data collection and evaluation methods. I found it hard to follow the content of the paper in its current state.
