연구성과
자료검색

연구자 및 연구성과 통합검색 KAIST-GIST-DGIST-UNIST 연구자, 연구실, 연구성과를 통합검색하실 수 있습니다.

검색 결과 내 재검색

검색어 | File Systems

연구자

1-3 out of 3 results

권영진 전산학부 2016 2, 2, 1, 1, 1, 2024
K
김문주 전산학부 2002 2, 1, 2, 1, 1, 2, 1, 3, 6, 1, 3, 2, 2, 3, 2, 1, 2, 1, 1, 1, 2025
K
원유집 전기및전자공학부 1999 2, 2, 2, 10, 3, 5, 4, 7, 3, 6, 6, 6, 3, 2, 6, 6, 6, 1, 4, 6, 3, 2025
K

연구실

검색결과가 없습니다.

저널논문

1-3 out of 7 results

TxFS: Leveraging File-system Crash Consistency to Provide ACID Transactions ACM TRANSACTIONS ON STORAGE, v.15, no.2, 2019-06
SCOPUS 15

Keywords Operating systems file systems crash consistency ACID transactions

TxFS: Leveraging File-system Crash Consistency to Provide ACID Transactions
Hu, Yige ; Zhu, Zhiting ; Neal, Ian ; Kwon, Youngjin ; Cheng, Tianyu ; Chidambaram, Vijay ; Witchel, Emmett

초록	We introduce TxFS, a transactional file system that builds upon a file system's atomic-update mechanism such as journaling. Though prior work has explored a number of transactional file systems, TxFS has a unique set of properties: a simple API, portability across different hardware, high performance, low complexity (by building on the file-system journal), and full ACID transactions. We port SQLite, OpenLDAP, and Git to use TxFS and experimentally show that TxFS provides strong crash consistency while providing equal or better performance.
서지사항	ACM TRANSACTIONS ON STORAGE, v.15, no.2, 2019-06
ISSN	1553-3077
주제분야	Computer Science
저자 키워드	Operating systems file systems crash consistency ACID transactions
DOI	10.1145/3318159
자료유형	Article; Proceedings Paper
기관	KAIST

Fair bandwidth allocating and strip-aware prefetching for concurrent read streams and striped RAIDs in distributed file systems JOURNAL OF SUPERCOMPUTING, v.74, no.8, pp.3904 - 3932, 2018-08
SCOPUS 4

Keywords Distributed file system RAID Concurrent read streams Fair bandwidth allocation Strip-aware prefetching

Fair bandwidth allocating and strip-aware prefetching for concurrent read streams and striped RAIDs in distributed file systems
Lee, Sangmin ; Hyun, Soon-Joo ; Kim, Hong-Yeon ; Kim, Young-Kyun

초록	With a striped RAID (Redundant Array of Independent Disks) which consists of multiple disks and spreads data across them in parallel, distributed file systems (DFSs) easily enhance the performance of a single read stream (i.e., a series of sequential reads by a process). However, most existing DFSs suffer from performance degradation in concurrent read streams (i.e., multiple series of sequential reads by concurrent processes). Furthermore, research on the performance of concurrent ones for a striped RAID in DFSs has been rarely reported so far. In this paper, we define the problems that degrade it at different configurations of striped RAIDs, and resolve them by proposing the following two methods: (1) a fair allocating of network bandwidth for concurrent read streams and (2) a strip-aware prefetching for each individual read stream. We show that our proposal outperforms all the existing DFSs by at least two times for all kinds and configurations of striped RAIDs. Furthermore, the performance gap between our proposal and the existing DFSs becomes wider according to the increasing number of striped disks.
서지사항	JOURNAL OF SUPERCOMPUTING, v.74, no.8, pp.3904 - 3932, 2018-08
ISSN	0920-8542
주제분야	Computer Science Engineering
저자 키워드	Distributed file system RAID Concurrent read streams Fair bandwidth allocation Strip-aware prefetching
DOI	10.1007/s11227-018-2396-4
자료유형	Article
기관	KAIST

APS: adaptable prefetching scheme to different running environments for concurrent read streams in distributed file systems JOURNAL OF SUPERCOMPUTING, v.74, no.6, pp.2870 - 2902, 2018-06
SCOPUS 12

Keywords Distributed file system Concurrent read streams Data prefetching Device type Network delay

APS: adaptable prefetching scheme to different running environments for concurrent read streams in distributed file systems
Lee, Sangmin ; Hyun, Soon-Joo ; Kim, Hong-Yeon ; Kim, Young-Kyun

초록	Distributed file systems (DFSs) are widely used in various areas. One of the key issues is to provide high performance of concurrent read streams (i.e., multiple series of sequential reads by concurrent processes) for their applications. Despite the many studies on local file systems (LFSs), research has seldom been done on concurrent read streams in DFSs with different running environments (i.e., different types of storage devices and various network delays). Furthermore, most of the existing DFSs have a sharply degraded performance compared with a LFS (i.e., EXT4). Therefore, to achieve high performance in concurrent read streams, this study introduces a populating effect that keeps sending subsequent reads to a storage server and then proposes an adaptable prefetching scheme (APS) to obtain the effect even in different running environments. Hence, our APS resolves all the problems that we identified as dramatically degrading the performance in existing DFSs. In three different types of storage devices and in various network delays, the evaluation results show that our prefetching scheme (1) achieves almost the same performance as a LFS from an individual server and (2) minimizes the performance degradation of random reads.
서지사항	JOURNAL OF SUPERCOMPUTING, v.74, no.6, pp.2870 - 2902, 2018-06
ISSN	0920-8542
주제분야	Computer Science Engineering
저자 키워드	Distributed file system Concurrent read streams Data prefetching Device type Network delay
DOI	10.1007/s11227-018-2333-6
자료유형	Article
기관	KAIST

연구과제

검색결과가 없습니다.

특허

1-2 out of 2 results

분산 파일 시스템에서 저장 공간 절약을 위한 효과적인 블록 인코딩 및 블록 배치 방법 An efficient block encoding and block placement method for reducing space overhead in distributed file systems, Registration Number: 10-1341386-0000, Registration Date: 2013-12-09, 대한민국

분산 파일 시스템에서 저장 공간 절약을 위한 효과적인 블록 인코딩 및 블록 배치 방법

요약정보	본 발명은 데이터 복제에 기반을 두어 내고장성을 지원하는 블록 단위 분산 파일 시스템에서 복제본의 저장에 따라 추가적으로 요구되는 저장 공간을 감소시키면서 동일 수준의 내고장성을 지원하는 공간 효율성을 개선하는 방법에 관한 것이다. 분산 파일 시스템의 데이터 노드들은 상기 적어도 하나의 파일을 구성하는 원본 데이터 블록들과 상기 원본 데이터 블록들이 복제된 복제 블록들을 저장한다. 저장 공간을 감소하기 위해 네임 노드는 상기 데이터 노드들의 블록 정보를 스캔하여 상기 데이터 노드들 각각에 저장된 데이터 블록들로 구성되는 적어도 하나의 스트라이프를 구성하고, 상기 데이터 노드들 각각은 상기 네임 노드에서 전달된 인코딩 명령을 실행하기 위하여 상기 데이터 블록들의 로컬 스토리지에서의 경로를 추출하고, 소거 코딩으로 인코딩하여 패리티 블록들을 생성하고, 상기 생성된 패리티 블록들 중 하나만을 자신의 데이터 노드에 저장하고 나머지 패리티 블록들은 다른 데이터 노드들에 시스템의 가용성을 보장하는 규칙에 따라 이동시켜 저장하고, 저장된 데이터 블록에 결함이 발생하는 경우, 상기 네임 노드로부터 상기 결함이 발생한 블록이 포함되는 스트라이프의 구성 정보를 수신하고 디코딩하여 상기 결함이 발생한 데이터 블록을 복구한다.	크게보기
타언어명	An efficient block encoding and block placement method for reducing space overhead in distributed file systems
출원번호	10-2012-0103555
출원일자	2012-09-18
등록번호	10-1341386-0000
등록일자	2013-12-09
발명자	이윤준, 이경하, 안후영, 김수형, 이상민, 김영균
출원/등록국가	대한민국
기관	KAIST

해쉬 트리 기반의 스트림 암호화 및 복호화 방법과 암호 파일 시스템 METHOD FOR ENCRYPTING AND DECRYPTING STREAM AND CRYPTOGRAPHIC FILE SYSTEMS THEREOF, Registration Number: 10-1133988-0000, Registration Date: 2012-03-30, 대한민국

해쉬 트리 기반의 스트림 암호화 및 복호화 방법과 암호 파일 시스템

요약정보	실시예는 스트림 암호화 방법에 관한 것이다.실시예에 따른 스트림 암호화 방법은, 파일 시스템 암호화에 사용할 루트 키를 생성하는 단계, 해쉬 함수를 통해 루트 키로부터 복수의 서브 키를 생성하여 키 트리를 구성하는 단계, 키 트리에 구성된 복수의 서브 키와 암호 파일 시스템의 스토리지에 구성된 데이터 블록들을 매핑하는 단계, 및 데이터 블록들에 매핑된 복수의 서브 키와, 평문 스트림을 XOR 연산하여 암호문 스트림을 생성하는 단계를 포함한다.	크게보기
타언어명	METHOD FOR ENCRYPTING AND DECRYPTING STREAM AND CRYPTOGRAPHIC FILE SYSTEMS THEREOF
출원번호	10-2010-0037773
출원일자	2010-04-23
등록번호	10-1133988-0000
등록일자	2012-03-30
발명자	박규호
출원/등록국가	대한민국
기관	KAIST

학술활동

1-3 out of 29 results

Breaking Barriers: Expanding GPU Memory with Sub-Two Digit Nanosecond Latency CXL Controller The 16th ACM Workshop on Hot Topics in Storage and File Systems, 미국, 2024-07-08 ~ 2024-07-09

Breaking Barriers: Expanding GPU Memory with Sub-Two Digit Nanosecond Latency CXL Controller
Gouk, Donghyun ; Kang, Seungkwan ; Bae, Hanyeoreum ; Ryu, Eojin ; Lee, Sangwon ; Kim, Dongpyung ; Jang, Junhyeok ; Jung, Myoungsoo

초록	This work introduces a GPU storage expansion solution utilizing CXL, featuring a novel GPU system design with multiple CXL root ports for integrating diverse storage media (DRAMs and/or SSDs). We developed and siliconized a custom CXL controller integrated at the hardware RTL level, achieving two-digit nanosecond roundtrip latency, the first in the field. This study also includes speculative read and deterministic store mechanisms to efficiently manage read and write operations to hide the endpoint's backend media latency variation. Performance evaluations reveal our approach significantly outperforms existing methods, marking a substantial advancement in GPU storage technology.
학술대회명	The 16th ACM Workshop on Hot Topics in Storage and File Systems
개최국	미국
개최장소	Santa Clara, CA
개최일	2024-07-09
저자 키워드
DOI
기관	KAIST

Advocating for Key-Value Stores with Workload Pattern Aware Dynamic Compaction USENIX Workshop on Hot Topics in Storage and File Systems, 미국, 2024-07-08 ~ 2024-07-09

Advocating for Key-Value Stores with Workload Pattern Aware Dynamic Compaction
Yoon, Heejin ; Yang, Jin ; Bang, Juyoung ; Noh, Sam H. ; Choi, Young-Ri

초록	In real life, the ratio of write and read operations of key-value (KV) store workloads usually changes over time. In this paper, we present a Dynamic wOrkload Pattern Aware LSM-based KV store (DOPA-DB), which supports dynamic compaction strategies depending on the workload pattern. In particular, DOPA-DB is a tiered LSM-based KV store with multiple key ranges, which enables varying compaction sizes. For write-intensive workloads, DOPA-DB can minimize write stalls while minimizing compaction overhead, and for read-intensive workloads, it can aggressively perform compaction to reduce the number of file accesses. Our preliminary experimental results show the potential benefits of dynamic compaction and provide insight into research directions for dynamic compaction strategies.
학술대회명	USENIX Workshop on Hot Topics in Storage and File Systems
개최국	미국
개최장소	Santa Clara
개최일	2024-07-08
저자 키워드
DOI	10.1145/3655038.3665955
기관	UNIST

SAND: A Storage Abstraction for Video-based Deep Learning HotStorage '23: 15th ACM Workshop on Hot Topics in Storage and File Systems, 미국, 2023-07-09 ~ 2023-07-09
Keywords computational storage storage abstraction video preprocessing

SAND: A Storage Abstraction for Video-based Deep Learning
Hong, Uitaek ; Lim, Hwijoon ; Yeo, Hyunho ; Park, Jinwoo ; Han, Dongsu

초록	Deep learning has gained significant success in video applications such as classification, analytics, and self-supervised learning. However, when scaling out to a large volume of videos, existing approaches suffer from a fundamental limitation; they cannot efficiently utilize GPUs for training deep neural networks (DNNs). This is because video decoding in data preparation incurs a prohibitive amount of computing overhead, making GPU idle for the majority of training time. Otherwise, caching raw videos in memory or storage to bypass decoding is not scalable as they account for from tens to hundreds of terabytes.This paper proposes SAND, a system that enables deep learning frameworks to directly access training data by a storage abstraction. This abstraction effectively hides the data preprocessing delay, enabling GPUs to be fully utilized for DNN training. To accomplish this, SAND operates an in-storage cache and manages the cache by ahead-of-time scheduling to guarantee that requested training data can be always retrieved immediately from the cache. This scheduling considers the future data accesses of deep learning frameworks for cache replacement. Compared to the existing approach, our evaluation using emulated environments shows that SAND improves the GPU utilization by 6.0X and reduces the training time by 75.9% on average.
학술대회명	HotStorage '23: 15th ACM Workshop on Hot Topics in Storage and File Systems
개최국	미국
개최장소	Boston, MA
개최일	2023-07-09
저자 키워드	computational storage storage abstraction video preprocessing
DOI	10.1145/3599691.3603407
기관	KAIST

연구장비

검색결과가 없습니다.