TomDB: The World's First Key-Value Store Database Using a Hardware Accelerator

- Data-Centric Computing and Systems Laboratory (DISCOS) -

▲ Professor Kim Young-jae, Department of Computer Science & Engineering

Professor Kim Young-jae of the Department of Computer Science & Engineering has developed TomDB, a hardware-accelerator-based key-value store database, in collaboration with SK hynix's Memory Systems Research and the US Department of Energy's Los Alamos National Laboratory.

The figure below shows the hardware and software design of TomDB. TomDB consists of a client component (a user-space device driver) and a server component (the TomDB engine). The server component is the database engine of TomDB and runs on a Fidus Sidewinder-100 (SW-100) FPGA development board.

▲ (Top left) The TomDB software stack; (bottom left) the architecture of an HPC platform adopting TomDB; (right) the TomDB prototype (exhibition photograph, Flash Memory Summit 2022)

TomDB accelerates attribute-based data retrieval in databases holding vast amounts of data, and shows higher data retrieval and query processing performance than the existing RocksDB.

The work has also been accepted for publication at the outstanding international conference IEEE Cluster 2023 (acceptance rate: 32/130 = 24.6%) under the title 'TomDB: A Hardware-Accelerated Key-Value Store for Data-Intensive Applications'.

IEEE Cluster is an internationally distinguished conference covering parallel computer (cluster) systems and the various parallel computing applications that use them. Cluster is listed as an outstanding conference in the BK21 list of outstanding international conferences in computer science and in the Korean Institute of Information Scientists and Engineers' 2020 list of outstanding conferences in the software field. This year it will be held in Santa Fe, New Mexico, United States, from 31 October to 3 November.

▶ This research is the result of close collaboration carried out while Professor Kim Young-jae worked at SK hynix's Memory Systems Research during his sabbatical year (the second semester of 2022 and the first semester of 2023).

▶ Conference link: https://clustercomp.org/2023/

▶ Data-Centric Computing and Systems Laboratory: https://discos.sogang.ac.kr