Although solid-state drives offer incredibly fast read and write speeds on the hardware side, their outdated interface with the host file system has created performance bottlenecks on the software side. To tackle this issue, scientists at DGIST, Korea, have created ‘KEVIN,’ which brings together a key–value file system and a specially designed solid-state drive. Their approach solves many problems of traditional file systems and will help provide faster more efficient computing environments.
In most computers, ‘files’ can be opened with a single click, but behind this click are a set of highly orchestrated processes that convert these files to and from the digital form. There is a complex interface between the file system software and the device where files are stored, and the read/write speeds that we ultimately perceive (or how long it takes to open and save files) are closely related to the inner workings of this interface.
Nowadays, the performance and price of solid-state drives (SSD) has made them the go-to option in many applications, replacing the traditional hard disk drive (HDD). However, the available interfaces between file systems and physical drives aren't able to keep up with the modern storage devices. In the legacy block interface, which is widely used today, the host file system has to ‘transform’ fixed-sized blocks of bits into the files and vice versa when we read or write data. Because of the incredible speeds of SSDs, keeping accurate track of the correspondence between these blocks and the abstraction of files has created a series of bottlenecks in the file system side. In other words, even if we were to keep improving the performance of SSDs, we might hardly notice an improvement in the actual read/write speeds of files.
Fortunately, at Daegu Gyeongbuk Institute of Science and Technology (DGIST), Korea, scientists led by Assistant Professor Sungjin Lee are looking for ways to tackle this issue. In their latest study, which was presented at the 15th USENIX Symposium on Operating Systems Design and Implementation, the team reported a strategy by which these bottlenecks can be overcome. Their approach is a novel take on a paradigm called key–value storage, by which a file is not managed as a combination of equally sized blocks, but as variable-length objects accessed by uniquely assigned keys.
They combined a file system (software) and a specially designed SSD (hardware) to create ‘KEVIN.’ The main benefit of KEVIN over previous key–value storage approaches is that it offloads many important functionalities of the file system directly into the storage device side. “Our strategy let us make the file system design much lighter and solved multiple issues with traditional file systems, achieving speedups of roughly 68% on average for realistic workloads,” remarks Dr Lee.
Overall, the study aptly illustrates the glaring problems of currently used file systems and offers an innovative solution. “File systems are a fundamental piece of system software that stores and manages all files from users and applications, and we envision that our research may provide a faster and more efficient computing environment for everyone,” highlights Dr Lee. It’s high time we move on from legacy interfaces and take full advantage of the capabilities of our hardware!
Jinhyung Koo1, Junsu Im1, Jooyoung Song1, Juhyung Park1, Eunji Lee2, Bryan S. Kim3, Sungjin Lee*1
Title of original paper:
Modernizing File System through In-Storage Indexing
15th USENIX Symposium on Operating Systems Design and Implementation
1Daegu Gyeongbuk Institute of Science and Technology (DGIST)
*Corresponding author’s email: [email protected]
About Daegu Gyeongbuk Institute of Science and Technology (DGIST)
Daegu Gyeongbuk Institute of Science and Technology (DGIST) is a well-known and respected research institute located in Daegu, Republic of Korea. Established in 2004 by the Korean Government, the main aim of DGIST is to promote national science and technology, as well as to boost the local economy.
With a vision of “Changing the world through convergence", DGIST has undertaken a wide range of research in various fields of science and technology. DGIST has embraced a multidisciplinary approach to research and undertaken intensive studies in some of today's most vital fields. DGIST also has state-of-the-art-infrastructure to enable cutting-edge research in materials science, robotics, cognitive sciences, and communication engineering.
About the author
Dr. Sungjin Lee is currently an Assistant Professor at DGIST. He received a BE degree in electrical engineering from Korea University in 2005 and MS and PhD degrees in computer science and engineering from Seoul National University, Korea, in 2007 and 2013, respectively. Before joining DGIST, he was a postdoctoral associate in the Computer Science and Artificial Intelligence Laboratory at the Massachusetts Institute of Technology, Cambridge, USA. His current research interests include storage systems, operating systems, and system software.