AI helps you reading Science

AI generates interpretation videos

AI extracts and analyses the key points of the paper to generate videos automatically


pub
Go Generating

AI Traceability

AI parses the academic lineage of this thesis


Master Reading Tree
Generate MRT

AI Insight

AI extracts a summary of this paper


Weibo:
Legacy file systems focus on sequential access optimization instead of the write amplification reduction, which is critical for flash memory

Extending the lifetime of flash-based storage through reducing write amplification from file systems

FAST, pp.257-270, (2013)

Cited by: 153|Views111
EI
Full Text
Bibtex
Weibo

Abstract

Flash memory has gained in popularity as storage devices for both enterprise and embedded systems because of its high performance, low energy and reduced cost. The endurance problem of flash memory, however, is still a challenge and is getting worse as storage density increases with the adoption of multi-level cells (MLC). Prior work has ...More

Code:

Data:

0
Introduction
  • Flash memory technology has greatly improved. As mainstream design moves from single level cell (SLC) to multi-/triple-level cells (MLC/TLC), flash-based storage is witnessing an increase of capacity and a reduction of per-bit cost, which results in a sharp growth of adoption in both enterprise and embedded storage systems.
  • File systems try to reduce updates with tail packing [7] or data compression [2, 13].
  • These data reduction techniques are inefficient in reducing write amplification from file systems, because the metadata updates generated from file systems can hardly be reduced for the following two reasons.
  • Flash blocks are further arranged into planes and channels for parallelism inside a flash drive, known as internal parallelism
Highlights
  • In recent years, flash memory technology has greatly improved
  • As mainstream design moves from single level cell (SLC) to multi-/triple-level cells (MLC/TLC), flash-based storage is witnessing an increase of capacity and a reduction of per-bit cost, which results in a sharp growth of adoption in both enterprise and embedded storage systems
  • We measure the write amplification, the total size/count of writes to the flash memory divided by the total size/count of writes issued from the application, to evaluate the write efficiency of Object Flash Storage System, as well as ext3, ext2, and btrfs on an up-to-date page-level flash translation layers
  • Legacy file systems focus on sequential access optimization instead of the write amplification reduction, which is critical for flash memory
  • We propose an object-based design name OFTL, in which the storage management is offloaded to the flash translation layers from the file system for direct management over the flash memory
  • Page metadata is used to keep the inverse index for the lazy indexing and the transaction information to provide write atomicity for journal removal, with the help of the updating window to track the latest allocated flash blocks that have not been checkpointed
Methods
  • Write and close operations (RWC Trace) from the system level IO traces and replay them on file systems and OFSS.
  • The close operation incurs a fsync operation in the file systems or an object flush operation in OFSS.
  • In OFSS, the authors collect the count and size of I/Os to the flash memory in the storage layer of OFTL.
  • The authors collect the count and size of I/Os to flash memory in PFTL.
Results
  • The authors measure the write amplification, the total size/count of writes to the flash memory divided by the total size/count of writes issued from the application, to evaluate the write efficiency of OFSS, as well as ext3, ext2, and btrfs on an up-to-date page-level FTL.

    the authors first evaluate the overall performance of these four systems and analyze the metadata amplification.
  • The authors measure the write amplification, the total size/count of writes to the flash memory divided by the total size/count of writes issued from the application, to evaluate the write efficiency of OFSS, as well as ext3, ext2, and btrfs on an up-to-date page-level FTL.
  • The authors first evaluate the overall performance of these four systems and analyze the metadata amplification.
  • The authors measure the impact of flash page sizes and the overhead of extending the updating window brought by the OFTL design
Conclusion
  • OFSS brings larger benefits for workloads that have a large number of page unaligned updates or those that have frequent data synchronization.
  • The improvement of TPC-C workload from OFSS is not as significant as that of the other workloads.Legacy file systems focus on sequential access optimization instead of the write amplification reduction, which is critical for flash memory
  • System mechanisms, such as journaling, metadata synchronization and page-aligned updates, tremendously amplify the write intensity, while transparency brought by the indirection prevents the system from exploiting the flash memory characteristics.
  • With the system co-design with flash memory, write amplification from file systems is significantly reduced
Tables
  • Table1: Object Interface
  • Table2: Characteristics of the Workloads workload
  • Table3: Overall Evaluation on Write Amplification
Download tables as Excel
Related work
  • Flash file systems [32, 10, 9] have leveraged the nooverwrite property of flash memory for log-structured updates to optimize the performance by directly manag-

    11 11th USENIX Conference on File and Storage Technologies (FAST ’13) 267 ing the flash memory. But they have not done much to the flash endurance other than wear leveling. Nameless Writes [35] and Direct File System [20] propose to remove the mappings of the file system to allow the FTL to manage the storage space in order to improve the performance. But file semantics fail to be passed to the FTL for intelligent storage space management. OFTL takes the concept of object-based storage [25] to export the object interfaces to file systems, which easily passes the file semantics to the device for intelligent data layout.

    Recent research has also proposed eliminating the duplicated writes caused by the journaling in the flash storage. Write Atomic [28] exports an atomic write interface by leveraging the log-based structure of VSL [4] in FusionIO ioDrives. TxFlash [29] uses a cyclic commit protocol with the help of the page metadata to export the same interface. However, the cyclic property is complex to maintain, and the protocol requires a whole-drive scan after failures. Comparatively, OFTL takes a simple protocol similar to the transaction support in the log-structured file system [31], keeps the transaction information in the page metadata, and tracks the recently updated flash blocks in the updating window for fast recovery.
Funding
  • This work is supported by the National Natural Science Foundation of China (Grant No 60925006), the State Key Program of National Natural Science of China (Grant No 61232003), the National High Technology Research and Development Program of China (Grant No 2013AA013201), and the research fund of Tsinghua-Tencent Joint Laboratory for Internet Innovation Technology
Reference
  • blktrace(8) - linux man page. http://linux.die.net/man/8/blktrace.
    Findings
  • Btrfs. http://btrfs.wiki.kernel.org.
    Findings
  • Dbt2 test suite. http://sourceforge.net/apps/mediawiki/osdldbt.
    Findings
  • Fusionio virtual storage layer. http://www.fusionio.com/products/vsl.
    Findings
  • Lasr system call io trace. http://iotta.snia.org/tracetypes/1.
    Findings
  • The nvm express standard. nvmexpress.org.
    Google ScholarFindings
  • Reiserfs. http://reiser4.wiki.kernel.org.
    Findings
  • strace(1) - linux man page. http://linux.die.net/man/1/strace.
    Findings
  • Ubifs - ubi file-system. http://www.linux-mtd.infradead.org/doc/ubifs.html.
    Findings
  • Yaffs. http://www.yaffs.net.
    Findings
  • N. Agrawal, V. Prabhakaran, T. Wobber, J. D. Davis, M. Manasse, and R. Panigrahy. Design tradeoffs for ssd performance. In Proceedings of USENIX 2008 Annual Technical Conference, 2008.
    Google ScholarLocate open access versionFindings
  • S. Boboila and P. Desnoyers. Write endurance in flash drives: measurements and analysis. In Proceedings of the 8th USENIX conference on File and storage technologies (FAST), 2010.
    Google ScholarLocate open access versionFindings
  • Michael Burrows, Charles Jerian, Butler Lampson, and Timothy Mann. On-line data compression in a log-structured file system. In Proceedings of the fifth international conference on Architectural support for programming languages and operating systems (ASPLOS), 1992.
    Google ScholarLocate open access versionFindings
  • L.P. Chang. On efficient wear leveling for large-scale flash-memory storage systems. In Proceedings of the 2007 ACM symposium on Applied computing, 2007.
    Google ScholarLocate open access versionFindings
  • F. Chen, T. Luo, and X. Zhang. Caftl: A contentaware flash translation layer enhancing the lifespan of flash memory based solid state drives. In Proceedings of the 9th USENIX conference on File and Storage Technologies (FAST), 2011.
    Google ScholarLocate open access versionFindings
  • V. Chidambaram, T. Sharma, A. C. ArpaciDusseau, and R. H. Arpaci-Dusseau. Consistency without ordering. In Proceedings of the 10th USENIX conference on File and Storage Technologies (FAST), 2012.
    Google ScholarLocate open access versionFindings
  • L. M. Grupp, J. D. Davis, and S. Swanson. The bleak future of nand flash memory. In Proceedings of the 10th USENIX conference on File and Storage Technologies (FAST), 2012.
    Google ScholarLocate open access versionFindings
  • L.M. Grupp, A.M. Caulfield, J. Coburn, S. Swanson, E. Yaakobi, P.H. Siegel, and J.K. Wolf. Characterizing flash memory: anomalies, observations, and applications. In Proceedings of the 42nd Annual IEEE/ACM International Symposium on Microarchitecture (MICRO-42), 2009.
    Google ScholarLocate open access versionFindings
  • T. Harter, C. Dragga, M. Vaughn, A. C. ArpaciDusseau, and R. H. Arpaci-Dusseau. A file is not a file: understanding the i/o behavior of apple desktop applications. In Proceedings of the 23rd ACM Symposium on Operating Systems Principles (SOSP), 2011.
    Google ScholarLocate open access versionFindings
  • W. K. Josephson, L. A. Bongo, D. Flynn, and K. Li. Dfs: a file system for virtualized flash storage. In Proceedings of the 8th USENIX conference on File and Storage Technologies (FAST), 2010.
    Google ScholarLocate open access versionFindings
  • Y. Kang, J. Yang, and E. L. Miller. Object-based scm: An efficient interface for storage class memories. In Proceedings of IEEE 27th Symposium on Mass Storage Systems and Technologies (MSST), 2011.
    Google ScholarLocate open access versionFindings
  • S. W. Lee and B. Moon. Design of flash-based dbms: an in-page logging approach. In Proceedings of the 2007 ACM SIGMOD International Conference on Management of Data (SIGMOD), 2007.
    Google ScholarLocate open access versionFindings
  • D. Ma, J. Feng, and G. Li. Lazyftl: A pagelevel flash translation layer optimized for nand flash memory. In Proceedings of the 2011 International Conference on Management of Data (SIGMOD), 2011.
    Google ScholarLocate open access versionFindings
  • P. Macko, M. Seltzer, and K.A. Smith. Tracking back references in a write-anywhere file system. In Proceedings of the 8th USENIX conference on File and storage technologies (FAST), 2010.
    Google ScholarLocate open access versionFindings
  • M. Mesnier, G.R. Ganger, and E. Riedel. Objectbased storage. IEEE Communications Magazine, 41(8):84–90, 2003.
    Google ScholarLocate open access versionFindings
  • D. Nellans, M. Zappe, J. Axboe, and D. Flynn. ptrim ()+ exists (): Exposing new ftl primitives to applications. In 2nd Annual Non-Volatile Memory Workshop, 2011.
    Google ScholarFindings
  • P. O’Neil, E. Cheng, D. Gawlick, and E. O’Neil. The log-structured merge-tree (lsm-tree). Acta Informatica, 33(4):351–385, 1996.
    Google ScholarLocate open access versionFindings
  • X. Ouyang, D. Nellans, R. Wipfel, D. Flynn, and D. K. Panda. Beyond block i/o: Rethinking traditional storage primitives. In Proceedings of the 17th IEEE International Symposium on High Performance Computer Architecture (HPCA), 2011.
    Google ScholarLocate open access versionFindings
  • V. Prabhakaran, T. L. Rodeheffer, and L. Zhou. Transactional flash. In Proceedings of the 8th USENIX conference on Operating Systems Design and Implementation (OSDI), 2008.
    Google ScholarLocate open access versionFindings
  • O. Rodeh, J. Bacik, and C. Mason. Btrfs: The linux b-tree filesystem. Technical report, IBM Almaden Reserach Center, 2012.
    Google ScholarFindings
  • M. Seltzer. Transaction support in a logstructured file system. In Proceedings of the Ninth International Conference on Data Engineering, 1993.
    Google ScholarLocate open access versionFindings
  • David Woodhouse. Jffs2: The journalling flash file system, version 2. http://sourceware.org/jffs2.
    Findings
  • G. Wu and X. He. Delta-ftl: improving ssd lifetime via exploiting content locality. In Proceedings of the 7th ACM European Conference on Computer Systems (EuroSys), 2012.
    Google ScholarLocate open access versionFindings
  • Q. Yang and J. Ren. I-cash: Intelligently coupled array of ssd and hdd. In Proceedings of the 17th IEEE International Symposium on High Performance Computer Architecture (HPCA), 2011.
    Google ScholarLocate open access versionFindings
  • Y. Zhang, L. P. Arulraj, A. C. Arpaci-Dusseau, and R. H. Arpaci-Dusseau. De-indirection for flashbased ssds with nameless writes. In Proceedings of the 10th USENIX conference on File and Storage Technologies (FAST), 2012.
    Google ScholarLocate open access versionFindings
Your rating :
0

 

Tags
Comments
数据免责声明
页面数据均来自互联网公开来源、合作出版商和通过AI技术自动分析结果,我们不对页面数据的有效性、准确性、正确性、可靠性、完整性和及时性做出任何承诺和保证。若有疑问,可以通过电子邮件方式联系我们:report@aminer.cn
小科