A paper by Lee Chang-gyu (first author), a combined master's–doctoral student of the Laboratory for Advanced System Software (supervisor: Professor Kim Young-jae), has been accepted at the IEEE MASCOTS 2019 international conference (recognized IF: 2), which is listed among the outstanding international conferences in computer science under the BK21 Plus programme.

▲ Graduate student Lee Chang-gyu

(Combined master's–doctoral student; supervisor: Professor Kim Young-jae)

A paper by Lee Chang-gyu, a combined-programme graduate student of the Laboratory for Advanced System Software (LASS) (supervisor: Professor Kim Young-jae), has been accepted at the 2019 IEEE International Symposium on the Modeling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS). MASCOTS is a conference on the modelling, analysis and simulation of computer systems — storage and file systems, energy, networks, computer architecture and more — and is listed as an outstanding conference in the KIISE 'Revised List of Outstanding Software Conferences 2018' (see attachment 1). It is also listed at a recognized IF of 2 in the 'List of Outstanding International Conferences in Computer Science under the BK21 Plus Programme' (see attachment 2).

The paper is titled 'iLSM-SSD: An Intelligent LSM-tree based Key-Value SSD for Data Analytics' (first author: Lee Chang-gyu; corresponding author: Kim Young-jae; co-authors: Kang Hyun-gu, Park Dong-gyu, Park Sung-yong, Noh Jung-ki, Jung Woo-seok, Park Kyung). The paper proposes iLSM-SSD, which provides inside the SSD the key-value store widely used in distributed databases for large-scale processing. iLSM-SSD is based on the log-structured merge-tree (LSM-tree) widely used in key-value stores, with a design that takes account of the relatively limited computing and memory resources of an SSD. It also provides near-data processing, minimizing data movement between the SSD and the host machine by performing computation inside the SSD.

A key-value store provides a concise interface in which data (the value) can be accessed using only the key. Because of this concise interface, it is used as the backbone for data storage in systems such as distributed NoSQL databases for large-scale data processing. By providing the key-value store inside the SSD, iLSM-SSD removes the performance degradation caused by operating system layers such as the file system. It is therefore expected to deliver higher I/O efficiency in the various large-scale data processing systems that depend on key-value stores, and to free more computing resources for computation.

The research was conducted in collaboration with Sogang University's Distributed Cloud Laboratory (co-author: Park Dong-gyu; supervisor: Professor Park Sung-yong) and SK hynix (co-authors: Noh Jung-ki, Jung Woo-seok, Park Kyung).