With the emerging of the big data era, the scalability
problem of relational database management system (RDBMS)
becomes more and more severe. Thus NoSQL key-value database
is considered as a promising candidate for replacing RDBMS.
Aiming at removing its unnecessary overhead, key-value-based
hard disk drive (HDD) has been proposed to achieve higher
performance. However, the poor performance of HDD becomes
the bottleneck. As the cost per gigabyte of flash memory is
getting closer to HDD, high performance solid state drive (SSD)
is regarded as the best substitute of HDD. Nevertheless, there
are still some issues that needs to be taken into consideration.
In [1], Chen et al. have reported that applying key-value store to
SSD with conventional FTL would incur internal fragmentation
and further cause the degradation of device lifespan. Although
they proposed KV-FTL to deal with the above issues, their work
gave rise to the read amplification problem. In this paper, we
investigate the root cause of the read amplification problem and
propose the EMT-FTL to mitigate internal fragmentation and
read amplification with acceptable memory overhead. Compared
with KV-FTL, the experimental results showed that the proposed
EMT-FTL could improve overall performance by 17% with a
negligible loss in space utilization.

With the emerging of the big data era, the scalability
problem of relational database management system (RDBMS)
becomes more and more severe. Thus NoSQL key-value database
is considered as a promising candidate for replacing RDBMS.
Aiming at removing its unnecessary overhead, key-value-based
hard disk drive (HDD) has been proposed to achieve higher
performance. However, the poor performance of HDD becomes
the bottleneck. As the cost per gigabyte of flash memory is
getting closer to HDD, high performance solid state drive (SSD)
is regarded as the best substitute of HDD. Nevertheless, there
are still some issues that needs to be taken into consideration.
In [1], Chen et al. have reported that applying key-value store to
SSD with conventional FTL would incur internal fragmentation
and further cause the degradation of device lifespan. Although
they proposed KV-FTL to deal with the above issues, their work
gave rise to the read amplification problem. In this paper, we
investigate the root cause of the read amplification problem and
propose the EMT-FTL to mitigate internal fragmentation and
read amplification with acceptable memory overhead. Compared
with KV-FTL, the experimental results showed that the proposed
EMT-FTL could improve overall performance by 17% with a
negligible loss in space utilization.

[1] Yen-Ting Chen, Ming-Chang Yang, Yuan-Hao Chang, Tseng-Yi Chen,
Hsin-Wen Wei, and Wei-Kuan Shih. Co-Optimizing Storage Space Uti-
lization and Performance for Key-Value Solid State Drives. IEEE Trans-
actions on Computer-Aided Design of Integrated Circuits and Systems,
38:29{42, January 2019.
[2] Yang Seok Ki. Key Value SSD Explained - Concept, Device, System, and
Standard. https://www.snia.org/sites/default/files/SDC/2017/
presentations/Object_ObjectDriveStorage/Ki_Yang_Seok_Key_
Value_SSD_Explained_Concept_Device_System_and_Standard.pdf,
February 2017.
[3] Hitachi Accelerated Flash 2.0 an Innovative Approach to Solid-State
Storage. https://community.hds.com/docs/DOC-1005695, September
2018.
[4] Brian F. Cooper, Adam Silberstein, Erwin Tam, Raghu Ramakrishnan,
and Russell Sears. Benchmarking Cloud Serving Systems with YCSB. In
SoCC '10: Proceedings of the 1st ACM symposium on Cloud computing,
pages 143{154, June 2010.
[5] Nawsher Khan, Ibrar Yaqoob, Ibrahim Abaker Targio Hashem, Zakira
Inayat, Waleed Kamaleldin, Muhammad Alam, Muhammad Shiraz, and
Abdullah Gani. Big Data: Survey, Technologies, Opportunities, and
Challenges. The Scientic World Journal, page 18, July 2014.
[6] Sanjay Ghemawat and Je Dean. LevelDB Library Documentation.
https://github.com/google/leveldb.
[7] Kinetic HDD Repository. http://www.seagate.com/products/
enterprise-servers-storage/nearline-storage/kinetic-hdd/,
November 2015.
[8] Jim Elliott and Bob Brennan. Industry Innovation with Samsung's Next
Generation V-NAND. In Flash Memory Summit 2014, August 2014.
[9] Lanyue Lu, Thanumalayan Sankaranarayana Pillai, Andrea C. Arpaci-
Dusseau, and Remzi H. Arpaci-Dusseau. WiscKey: Separating Keys
from Values in SSD-conscious Storage. 14th USENIX Conference on
File and Storage Technologies FAST 16, pages 133{148, February 2016.
[10] Sung-Ming Wu, Kai-Hsiang Lin, and Li-Pin Chang. KVSSD: Close
Integration of LSM Trees and Flash Translation Layer forWrite-Ecient
KV Store. In 2018 Design, Automation and Test in Europe Conference
and Exhibition (DATE), pages 563{568, March 2018.
[11] Juan Li, Zhengguo Chen, Zhiguang Chen, Nong Xiao, Fang Liu, andWei
Chen. KV-FTL: A Novel Key-Value-Based FTL Scheme for Large Scale
SSDs. In 2017 IEEE 19th International Conference on High Perfor-
mance Computing and Communications; IEEE 15th International Con-
ference on Smart City; IEEE 3rd International Conference on Data Sci-
ence and Systems (HPCC/SmartCity/DSS), pages 106{114, December
2017.
[12] KV-SSD Seminar. https://github.com/OpenMPDK/KVSSD/wiki/
presentation/kvssd_seminar_2018/kvssd_seminar_2018_fw_
introduction.pdf, 2018.
[13] Berk Atikoglu, Yuehai Xu, and Eitan Frachtenberg. Workload Analysis
of a Large-Scale Key-Value Store. ACM SIGMETRICS Performance
Evaluation Review, 40, June 2012.