AI帮你理解科学

AI 生成解读视频

AI抽取解析论文重点内容自动生成视频


pub
生成解读视频

AI 溯源

AI解析本论文相关学术脉络


Master Reading Tree
生成 溯源树

AI 精读

AI抽取本论文的概要总结


微博一下
In this paper we have examined three alternatives for file storage on mobile computers: a magnetic disk, a flash disk emulator, and a flash memory card

Storage Alternatives for Mobile Computers

OSDI, pp.25-37, (1994)

引用381|浏览36
EI
下载 PDF 全文
引用
微博一下

摘要

Mobile computers such as notebooks, subnotebooks, and palmtops require low weight, low power consumption, and good interactive performance. These requirements impose many challenges on architectures and operating systems. This paper investigates three alternative storage devices for mobile computers: magnetic hard disks, flash memory disk...更多

代码

数据

0
简介
  • Mobile computer environments are different from traditional workstations because they require light-weight, low-

    This work was performed at Panasonic Technologies, Inc.’s Matsushita Information Technology Laboratory.

    Current storage technologies offer two alternatives for file storage on mobile computers: magnetic hard disks and flash memory.
  • The main disadvantages of flash memory are that it costs more than disks—30– 50/Mbyte, compared to 1–5/Mbyte for magnetic disks— and that it requires erasing before it can be overwritten.
  • It comes in two forms: flash memory cards and flash disk emulators.1.
  • It comes in two forms: flash memory cards and flash disk emulators.1 These devices behave differently, having varying access times and bandwidths
重点内容
  • Mobile computer environments are different from traditional workstations because they require light-weight, low-

    This work was performed at Panasonic Technologies, Inc.’s Matsushita Information Technology Laboratory.

    Current storage technologies offer two alternatives for file storage on mobile computers: magnetic hard disks and flash memory
  • The three basic storage architectures we studied are magnetic disks, flash disk emulators, and flash memory cards
  • In this paper we have examined three alternatives for file storage on mobile computers: a magnetic disk, a flash disk emulator, and a flash memory card
  • We have shown that either form of flash memory is an attractive alternative to magnetic disk for file storage on mobile computers
  • The power management of a disk file system spins down the disk when it is idle
  • At a utilization of 95%, compared to 40% utilization, the energy consumption rises by up to 150%, while the average write time increases up to 30%
  • Even with power management, a disk file system can consume an order of magnitude more energy than a file system using flash memory
结果
  • The authors used the simulator to explore the architectural tradeoffs between disks, flash disks, and flash cards.
  • 5.1 Basic Comparisons Tables 4(a)–(c) show for three traces and each device the energy consumed, and the average, mean, and standard deviations of the read and write response times.
  • SDP10 SDP5 Intel flash card Intel flash card measured datasheet datasheet measured datasheet measured datasheet Energy (J).
  • Read Response Mean Max. Write Response Mean Max (a) MAC trace measured datasheet datasheet measured datasheet measured datasheet.
结论
  • In this paper the authors have examined three alternatives for file storage on mobile computers: a magnetic disk, a flash disk emulator, and a flash memory card.
  • The authors have shown that either form of flash memory is an attractive alternative to magnetic disk for file storage on mobile computers.
  • The main disadvantage of using magnetic disk for file storage on mobile computers is its great energy consumption.
  • The power management of a disk file system spins down the disk when it is idle.
  • Even with power management, a disk file system can consume an order of magnitude more energy than a file system using flash memory
表格
  • Table1: Measured performance of three storage devices on an HP OmniBook 300
  • Table2: Manufacturers’ specifications for three storage devices. Latency for read/write operations indicates the overhead from a random operation, excluding the transfer itself (i.e., controller overhead, seeking, or rotational latency). The Intel erasure cost refers to a separate operation that takes 1.6s to erase 64 or 128 Kbytes (in this case latency and throughput are analogous)
  • Table3: Summary of (non-synthetic) trace characteristics. The statistics apply to the 90% of each trace that is actually simulated after the warm start. Note that it is not appropriate to compare performance or energy consumption of simulations of different traces, because of the different mean transfer sizes and durations of each trace
  • Table4: Comparison of energy consumption and response time for different devices, using the MAC, DOS, and HP traces. There was a 2-Mbyte DRAM buffer for MAC and DOS but no DRAM buffer cache in the HP simulations. Disk simulations spun down the disk after 5s of inactivity. Flash simulations were done with flash memory 80% utilized
Download tables as Excel
相关工作
  • In addition to the specific work on flash file systems mentioned previously, the research community has begun to explore the use of flash memory as a substitute for, or an ad-

    dition to, magnetic disks. Caceres et al proposed operating system techniques for exploiting the superior read performance of flash memory while hiding its poor write performance, particularly in a portable computer where all of DRAM is battery-backed [2]. Wu and Zwaenepoel discussed how to implement and manage a large non-volatile storage system, called eNVy, composed of NVRAM and flash memory for high-performance transaction processing. They simulated a system with Gbytes of flash memory and Mbytes of battery-backed SRAM, showing it could support the I/O corresponding to 30,000 transactions per second using the TPC-A database benchmark [23]. They found that at a utilization of 80%, 45% of the time is spent erasing or copying data within flash, while performance was severely degraded at higher utilizations [24]. Marsh et al examined the use of flash memory as a cache for disk blocks to avoid accessing the magnetic disk, thus allowing the disk to be spun down more of the time [15]. SunDisk recently performed a competitive analysis of several types of flash memory on an HP Omnibook 300 and found that the SunDisk SDP5-10 flash disk emulator was nearly an order of magnitude faster than an Intel Flash card using version 2 of the Flash Files System [22]. They also found that performance of the Intel Flash card degraded by 40% as it filled with data, with the most noticeable degradation between 95% and 99% storage utilization.
引用论文
  • Mary Baker, Satoshi Asami, Etienne Deprit, John Ousterhout, and Margo Seltzer. Non-volatile memory for fast, reliable file systems. In Proceedings of the Fifth International Conference on Architectural Support for Programming Languages and Operating Systems, pages 10–22, Boston, MA, October 1992. ACM.
    Google ScholarLocate open access versionFindings
  • Ramon Caceres, Fred Douglis, Kai Li, and Brian Marsh. Operating Systems Implications of Solid-State Mobile Computers. In Proceedings of the Fourth Workshop on Workstation Operating Systems, pages 21–27, Napa, CA, October 1993. IEEE.
    Google ScholarLocate open access versionFindings
  • Jeff Craig, March 1994. Personal communication.
    Google ScholarFindings
  • Brian Dipert and Markus Levy. Designing with Flash Memory. Annabooks, 1993.
    Google ScholarLocate open access versionFindings
  • Fred Douglis, P. Krishnan, and Brian Marsh. Thwarting the Power Hungry Disk. In Proceedings of 1994 Winter USENIX Conference, pages 293–306, San Francisco, CA, January 1994.
    Google ScholarLocate open access versionFindings
  • Hewlett-Packard. HP 100 and OmniBook Flash Disk Card User’s Guide, 1993.
    Google ScholarFindings
  • Hewlett-Packard. Kittyhawk HP C3013A/C3014A Personal Storage Modules Technical Reference Manual, March 1993. HP Part No. 5961-4343.
    Google ScholarLocate open access versionFindings
  • Intel. Mobile Computer Products, 1993.
    Google ScholarFindings
  • Intel. Flash Memory, 1994.
    Google ScholarFindings
  • Atsuo Kawaguchi, Shingo Nishioka, and Hiroshi Motoda. A flash-memory based file system. In Proceedings of the USENIX 1995 Winter Conference, New Orleans, January 1995. To appear.
    Google ScholarLocate open access versionFindings
  • Markus Levy. Interfacing Microsoft’s Flash File System. In Memory Products, pages 4–318–4–325. Intel Corp., 1993.
    Google ScholarFindings
  • Kester Li. Towards a low power file system. Technical Report UCB/CSD 94/814, University of California, Berkeley, CA, May 1994. Masters Thesis.
    Google ScholarFindings
  • Kester Li, Roger Kumpf, Paul Horton, and Thomas Anderson. A Quantitative Analysis of Disk Drive Power Management in Portable Computers. In Proceedings of the 1994 Winter USENIX, pages 279–291, San Francisco, CA, 1994.
    Google ScholarLocate open access versionFindings
  • B. Marsh and B. Zenel. Power Measurements of Typical Notebook Computers. Technical Report 110-94, Matsushita Information Technology Laboratory, May 1994.
    Google ScholarFindings
  • Brian Marsh, Fred Douglis, and P. Krishnan. Flash Memory File Caching for Mobile Computers. In Proceedings of the 27th Hawaii Conference on Systems Sciences, pages 451– 460, Maui, HI, 1994. IEEE.
    Google ScholarLocate open access versionFindings
  • Marshall Kirk McKusick, Michael J. Karels, and Keith Bostic. A pageable memory based file system. In USENIX Conference Proceedings, pages 137–144, Anaheim, CA, Summer 1990. USENIX.
    Google ScholarLocate open access versionFindings
  • NEC. Memory Products Data Book, Volume 1: DRAMS, DRAM Modules, Video RAMS, 1993.
    Google ScholarFindings
  • NEC. Memory Products Data Book, Volume 2: SRAMS, ASMs, EEPROMs, 1993.
    Google ScholarFindings
  • Mendel Rosenblum and John Ousterhout. The design and implementation of a log-structured file system. ACM Transactions on Computer Systems, 10(1):26–52, February 1992. Also appears in Proceedings of the 13th Symposium on Operating Systems Principles, October 1991.
    Google ScholarLocate open access versionFindings
  • Chris Ruemmler and John Wilkes. UNIX disk access patterns. In Proceedings of the Winter 1993 USENIX Conference, pages 405–420, San Diego, January 1993.
    Google ScholarLocate open access versionFindings
  • SunDisk Corporation. SunDisk SDP Series OEM Manual, 1993.
    Google ScholarFindings
  • SunDisk Corporation, 3270 Jay Street, Santa Clara, CA 95054. Competitive Analysis 80-40-00002 Rev. 1.0, 1994.
    Google ScholarFindings
  • Transaction Processing Performance Council. TPC Benchmark A Standard Specification Rev 1.1.
    Google ScholarFindings
  • Michael Wu and Willy Zwaenepoel. eNVy: a Non-Volatile, main memory storage system. In Proceedings of the Sixth International Conference on Architectural Support for Programming Languagesand Operating Systems, San Jose, CA, October 1994. To appear.
    Google ScholarLocate open access versionFindings
0
您的评分 :

暂无评分

标签
评论
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