Development of an Ultra-Fast File System 43 Times Faster Than Existing Ones

DGIST Professor Seongjin Lee's Team Explores Applications in Big Data, Artificial Intelligence, and Autonomous Driving

[Asia Economy Reporter Kim Bong-su] A computer data storage system with performance 43 times faster than existing systems has been developed. It also addresses the drawback of computer file systems whose performance deteriorates over time.

Daegu Gyeongbuk Institute of Science and Technology (DGIST) announced on the 19th that Professor Lee Sung-jin's team from the Department of Information and Communication Convergence has developed a new key-value interface-based file system that fundamentally solves the problems of traditional computer data storage systems. The developed storage system boasts performance up to 43 times faster than existing systems by applying a key-value interface using in-storage indexing technology within the storage device.

A 'file system' is basic system software that stores and indexes data, used across all fields of computer systems such as big data processing and artificial intelligence. Existing file systems store data in fixed-size blocks within the storage medium, which causes system performance to decline over time.

The research team focused on the key-value storage interface. The key-value interface is designed to flexibly change the size of blocks to process data, enabling more diverse data processing operations than the fixed-size block interface. They implemented an improved key-value interface by applying in-storage indexing technology and succeeded in developing a new concept file system called 'KEVIN'.

KEVIN not only operates with much faster performance and significantly lower workload but also drastically reduces the complex processes previously required to store and retrieve files by applying in-storage indexing technology. It achieved up to 43 times faster file operation processing performance while maintaining the same level of data reliability as existing systems, and secured performance stability sufficient for stable data input/output even as the file system ages.

Professor Lee Sung-jin said, "This research demonstrates the possibility of resolving bottlenecks occurring during data input/output in ultra-high-speed storage devices and achieving more than 40 times performance improvement. In the future, this technology could bring innovation to data processing not only in big data and artificial intelligence but also in the field of storage device design."

The research results were published on the 14th in the 'USENIX Symposium on Operating Systems Design and Implementation (OSDI)', a prestigious conference in the field of computer systems.

This content was produced with the assistance of AI translation services.