1 / 32

Seagate ships billionth drive!

Computer Structures and Design Lecture 30 – Disks 2008-04-28. Lecturer SOE Dan Garcia. How much storage do YOU have?. The ST506 was their first drive in 1979 (shown here), weighed 5 lbs, cost $1,500 and held 5 MeB. Today’s drives hold 1 TB, and are available for less than $200.

gala
Télécharger la présentation

Seagate ships billionth drive!

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Computer Structures and DesignLecture 30 – Disks2008-04-28 Lecturer SOE Dan Garcia How much storage do YOU have? The ST506 was their first drive in 1979 (shown here), weighed 5 lbs, cost $1,500 and held 5 MeB. Today’s drives hold 1 TB, and are available for less than $200. Seagate ships billionth drive! 1b.seagatestorage.com

  2. Magnetic Disk – common I/O device Computer Keyboard, Mouse Devices Processor (active) Memory (passive) (where programs, data live when running) Input Disk, Network Control (“brain”) Output Datapath (“brawn”) Display, Printer

  3. Magnetic Disk – common I/O device • 一种计算机存储器 • 信息存储在旋转盘的磁性材料的表面 • 和录音机类似，但这里记录的是数字量而非模拟量 • 非易失性存储（Nonvolatile storage） • 掉电时数据依然存在. • 两种类型 • 软盘 – slower, less dense, removable. • 硬盘 (HDD) – faster, more dense, non-removable. • 在计算机系统中的作用 (硬盘驱动器): • 用于长期, 廉价地存储文件 • 主存的备份“Backup”. 大容量, 低价格, 在存储结构中处于较底的层次 (虚拟内存)

  4. { Platters (1-12) Photo of Disk Head, Arm, Actuator 轴Spindle Arm Head Actuator

  5. Sector Disk Device Terminology • 多个盘片, 信息以磁的方式记录在盘的两个表面(usually) • Bits记录在道上, 可进一步分为扇区(e.g., 512 Bytes) • Actuator moves head(end of arm) over track (“seek”), wait for sectorrotate under head, then read or write Inner Track Outer Track Arm Head Platter Actuator

  6. Disk Device Performance (1/2) • 磁盘反应时间 = Seek Time + Rotation Time + Transfer Time + Controller Overhead • Seek Time? 依赖于arm要移动几条道, actuator的速度 • Rotation Time? 依赖于磁盘旋转的速度, 扇区离头的距离 • Transfer Time? 依赖于磁盘的数据率 (带宽) (f(bitdensity,rpm)), 及所访问的数据大小 Inner Track Sector Outer Track Head Controller Arm Spindle Platter Actuator

  7. Disk Device Performance (2/2) • 头移动到扇区的平均距离? • 转一圈的一半时间 • 7200转/分 120 转/秒 • 1 转 = 1/120 秒 8.33 毫秒 • 1/2 转 (revolution)  4.17 ms • arm移动的平均道数? • 磁盘工业标准测试: • 对各种可能的道求出所有寻道距离，加起来/ 所有可能个数 • 假定平均寻道距离是随机的 • 磁盘cache的大小会极大影响性能! • Cache内建于磁盘系统, OS不知细节

  8. Data Rate: Inner vs. Outer Tracks • 为了简化, 最初每道具有相同个数的扇区 • 由于外道更长, 每寸的位数更少 • 由于竞争的需要对所有的道都保持较高的每寸位数 (BPI) (“位密度为常数”) •  每个磁盘更大容量 •  边缘处每道有更多的扇区 •  由于磁盘以常速旋转,外道有更快的数据率 • 外道带宽是内道的1.7倍!

  9. Disk Performance Model /Trends • 容量 : + 100% / year (2X / 1.0 yrs) • 随着时间的发展, 增长速度如此之快以至于盘片数减少了 (现在有些只用1片!) • 传输率 (带宽) : + 40%/yr (2X / 2 yrs) • 转速+寻道时间 : – 8%/yr (1/2 in 10 yrs) • 面密度 • 沿道记录的位数: Bits/Inch(BPI) • 每面的道数: Tracks/Inch(TPI) • 我们关心单位面积的位密度Bits/Inch2 • 称为面密度=BPI x TPI • “~120 Gb/In2 is longitudinal limit” • “230 Gb/In2 now with perpendicular” • GB/$: > 100%/year (2X / 1.0 yrs) • Fewer chips + areal density

  10. 性能 企业应用, 服务器 E.g., Seagate Cheetah 15K.6 3 Gb/s,Serial Attached SCSI,Fibre Channel 450 GB, 3.5-inch disk 4 disks, 8 heads 15,000 RPM 12-17 watts (idle-normal) 3.4 ms avg. seek 164 MB/s transfer rate 1.6 Million Hrs MTBF 5 year warrantee $1000 = $3.30 / GB 容量 主流, 家庭使用 E.g., Seagate Barracuda 7200.11 SATA 3Gb/s NCQ,SATA 1.5Gb/s NCQ 1 TB, 3.5-inch disk 4 disks, 8 heads 7,200 RPM 8-12 watts (idle-normal) ?? ms avg. seek (上次8.5) 115 MB/s transfer rate 0.75 Million Hrs MTBF 5 year warrantee $200 = $0.20 / GB State of the Art:Two camps (2008) These use Perpendicular Magnetic Recording (PMR)!! source: www.seagate.com

  11. 1 inch disk drive! • Hitachi 2007 release • 由iPods& 数码相机驱动 • 8GB, 5-10MB/s (higher?) • 42.8 x 36.4 x 5 mm • 垂直磁记录 (PMR) • FUNDAMENTAL new technique • Evolution from Longitudinal • Starting to hit physical limit due to superparamagnetism（超顺磁性） • They say 10x improvement www.hitachi.com/New/cnews/050405.html www.hitachigst.com/hdd/research/recording_head/pr/

  12. Where does Flash memory come in? • Microdrives and Flash memory (e.g., CompactFlash) are going head-to-head • 都不易失 (no power, data ok) • Flash优点: 更耐用 & 低功耗(无移动部件, 微硬盘需要上下旋转) • Flash限制: 写的次数有限 (围绕电荷存储装置的绝缘氧化层的磨损) • How does Flash memory work? • NMOS晶体管，在门和源/出口间有附加的导体来 and “捕获traps”电子. 有/无电子对应于1或0. en.wikipedia.org/wiki/Flash_memory

  13. en.wikipedia.org/wiki/Ipodwww.apple.com/ipod What does Apple put in its iPods? Toshiba flash1, 2GB Samsung flash4, 8GB Toshiba flash 8, 16, 32GB Toshiba 1.8-inch HDD80, 160GB shuffle, nano, classic, touch

  14. Use Arrays of Small Disks… • Katz and Patterson asked in 1987: • 可否使用更小的磁盘来缩小磁盘和CPU性能间的鸿沟? 传统: 4磁盘设计 3.5” 5.25” 10” 14” High End Low End 磁盘阵列: 1磁盘设计 3.5”

  15. Replace Small # of Large Disks with Large # of Small! (1988 Disks) IBM 3390K 20 GBytes 97 cu. ft. 3 KW 15 MB/s 600 I/Os/s 250 KHrs $250K x70 23 GBytes 11 cu. ft. 1 KW 120 MB/s 3900 I/Os/s ??? Hrs $150K IBM 3.5" 0061 320 MBytes 0.1 cu. ft. 11 W 1.5 MB/s 55 I/Os/s 50 KHrs $2K 容量 体积 功耗 数据率 I/O Rate MTTF Cost 9X 3X 8X 6X 磁盘阵列具有更高的性能, 每单位体积更高的容量, 单位功耗更高的容量, 可靠性?

  16. Array Reliability • 可靠性 – 是否有一个组件失效 • 测量方法：平均故障时间 (MTTF) • N个磁盘的可靠性= 单个磁盘的可靠性 ÷ N(假定故障是独立的) • 50,000 Hours ÷ 70 disks = 700 hour • 磁盘系统平均故障时间: 从6年降为1个月! • Disk arrays too unreliable to be useful!

  17. Redundant Arrays of (Inexpensive) Disks • 文件 “条状”分布于多个磁盘 • 冗余产生高数据可用性 • 可用性: 即使一些组件出现故障，仍能为用户提供服务 • 磁盘仍会出故障 • 内容通过存储在阵列中的冗余数据重建  由于存储冗余信息，容量减少  由于更新冗余信息，带宽减少

  18. Berkeley History, RAID-I • RAID-I (1989) • 由以下部件组成：带有128 MB内存的Sun 4/280工作站, 四个双路（dual-string） SCSI控制器, 28 5寸SCSI 磁盘及专用的磁盘条带化软件 • Today RAID is > tens billion dollar industry, 80% nonPC disks sold in RAIDs

  19. “RAID 0”: No redundancy = “AID” • 本例中有四个磁盘, 块组织方式 • 更快的访问速度，因为可同时从多个磁盘传输数据 This and next 5 slides from RAID.edu, http://www.acnc.com/04_01_00.htmlhttp://www.raid.com/04_00.htmlalso has a great tutorial

  20. RAID 1: Mirror data • 每个磁盘完全复制到其 “镜像” • 可以获得极高的可用性 • 写数据带宽减少: • 1 逻辑写 = 2 物理写 • 非常昂贵的解决方案: 100% 容量浪费

  21. RAID 3: Parity • 计算出各组的奇偶效验存储到P磁盘，以此来保护磁盘减少故障 • 逻辑上, 单个高容量, 高传输率磁盘 • 本例中，奇偶效验会多出25%容量，而 RAID 1 多出１００％，(5 disks vs. 8 disks)

  22. Inspiration for RAID 5 (RAID 4 block-striping) • Small writes (write to one disk): • 方案1: 读其他磁盘的数据, 产生新的和，然后写奇偶效验盘(访问所有磁盘) • 方案2: 由于P拥有旧的和,比较新、旧数据, 差值加到P: 1 逻辑写 = 2 物理读 + 2 物理写到两个磁盘 • 奇偶效验盘是Small writes的瓶颈: 例如要写A0, B1都要写P盘 A0 B0 C0 D0 P D1 P A1 B1 C1

  23. RAID 5: Rotated Parity, faster small writes • 由于校验盘交错，分别写成为可能 • 例如: 写A0, B1会用到 0, 1, 4, 5磁盘, 因此可以并行处理 • 仍然是1 small write = 4次物理磁盘访问 en.wikipedia.org/wiki/Redundant_array_of_independent_disks

  24. Peer Instruction • RAID 1 (mirror) and 5 (rotated parity) help with performance and availability • RAID 1 has higher cost than RAID 5 • Small writes on RAID 5 are slower than on RAID 1 ABC 0: FFF 1: FFT 2: FTF 3: FTT 4: TFF 5: TFT 6: TTF 7: TTT

  25. “And In conclusion…” • 磁盘仍然高速增长: 容量60%/yr, 带宽40%/yr, 寻道时间更短, 旋转加快, MB/$每年增长100%? • Designs to fit high volume form factor • PMR a fundamental new technology • breaks through barrier • RAID • Higher performance with more disk arms per $ • Adds option for small # of extra disks • Can nest RAID levels • Today RAID is > tens-billion dollar industry, 80% nonPC disks sold in RAIDs, started at Cal

  26. Bonus slides • These are extra slides that used to be included in lecture notes, but have been moved to this, the “bonus” area to serve as a supplement. • The slides will appear in the order they would have in the normal presentation Bonus

  27. BONUS : Hard Drives are Sealed. Why? • The closer the head to the disk, the smaller the “spot size” and thus the denser the recording. • Measured in Gbit/in2. ~60 is state of the art. • Disks are sealed to keep the dust out. • Heads are designed to “fly” at around 5-20nm above the surface of the disk. • 99.999% of the head/arm weight is supported by the air bearing force (air cushion) developed between the disk and the head.

  28. Historical Perspective • 容量和尺寸，对于市场的影响要比对性能的影响大 • 尺寸因素的变革: 1970s: 主机系统  14 寸直径盘 1980s: 小型机, 服务器  8”, 5.25” diameter disks Late 1980s/Early 1990s: • PCs  3.5 inch diameter disks • Laptops, notebooks  2.5 inch disks • Palmtops didn’t use disks, so 1.8 inch diameter disks didn’t make it • Early 2000s: • MP3 players  1 inch disks

  29. Early Disk History (IBM) Data density Mbit/sq. in. Capacity of unit shown Megabytes 1973: 1. 7 Mbit/sq. in 140 MBytes 1979: 7. 7 Mbit/sq. in 2,300 MBytes source: New York Times, 2/23/98, page C3, “Makers of disk drives crowd even more data into even smaller spaces”

  30. Early Disk History 1989: 63 Mbit/sq. in 60,000 MBytes 1997: 1450 Mbit/sq. in 1600 MBytes 1997: 3090 Mbit/sq. in 8100 MBytes source: New York Times, 2/23/98, page C3, “Makers of disk drives crowd even more data into even smaller spaces”

  31. Disk Performance Example • Calculate time to read 1 sector (512B) for Deskstar using advertised performance; sector is on outer track Disk latency = average seek time + average rotational delay + transfer time + controller overhead = 8.5 ms + 0.5 * 1/(7200 RPM) + 0.5 KB / (100 MB/s) + 0.1 ms = 8.5 ms + 0.5 /(7200 RPM/(60000ms/M)) + 0.5 KB / (100 KB/ms) + 0.1 ms = 8.5 + 4.17 + 0.005 + 0.1 ms = 12.77 ms • How many CPU clock cycles is this?

More Related