Computer Organization and Architecture

External Memory

Outline

• Magnetic Disk

• RAID

• Solid State Drives

• Optical Memory

• Magnetic Tape

Magnetic disk

• A disk is a nonmagnetic circular platter coated with magnetizable material

• Data are recorded to or read from the disk using the read/write head

• Hard disks can have one platter, or more

• Direct access storage

Access method

• 顺序存取：数据以线性的方式存放在存储介质中。读取数据必须按照顺序的方式，从当前的位置按照顺序移动到数据所在的位置进行读取。典型的顺序存取是磁带

• 直接存取：数据按块存储在介质中，并且每个块都有一个唯一的地址。存取时，先按照这个唯一地址到达所在的块，然后在块中，顺序搜索到数据。典型的直接存取是硬盘

• 随机存取：每个存取单元都有一个唯一的地址，通过寻址机制可以直接找到这个位置，不依赖于之前的存取操作所在的位置。典型的随机存取是内存

• 关联存取：关联存取是通过对字中的部分内容进行比较，如果匹配就进行存取操作。关联存取是对字的内容进行比较，不是地址寻址。典型的关联存取是cache

Read and write mechanism

• Recording & retrieval via conductive coil called a head

  • May be single read/write head or separate ones

  • During read/write, head is stationary, platter rotates

• Write

  • Current through coil produces magnetic field

  • Pulses sent to head

  • Magnetic pattern recorded on surface below

• Read (traditional)

  • 磁头上有线圈，读的时候磁头不动，磁盘通过马达在旋转

  • 磁盘上的磁性物质被磁化成两个方向，表示0和1

  • 磁性物质在转动的时候，会在磁头的线圈上产生电流

  • 不同的磁化方向产生的电流方向不一样

  • 通过分析电流方向，就可以得到磁化方向，从而确定存储的是0还是1

  • Magnetic field moving relative to coil produces current

  • Coil is the same for read and write

  • Slow reading and writing speed

• Read (contemporary)

  • Separate read head, close to write head

  • MR（magneto resistive）： it senses the magnetization direction of upper disc

  • magnetization direction changes the resistance of the magneto resistive sensor

  • magnetization direction can be obtained by measuring the resistance value to determine the stored information

  • Set shield to prevent interference

  • High frequency operation，Higher storage density and speed

Terminology

• A hard disk may have several platters, with information recorded magnetically on both surfaces

Data Organization and Formatting

• Disk is divided into several concentric rings

  • Called track

  • Data is stored in the track

• tracks

  • Gaps between tracks to prevent interference between different tracks

  • Reduce gap to increase capacity

  • Disc rotates at constant angular velocity

  • Same number of bits per track (variable packing density)

• Tracks divided into sectors

  • The smallest unit of data storage

  • Each track generally contains several hundred sectors

  • Current sector size is 512 bytes

  • Data is written to or read from the disk in sectors

• May have more than one sector per block

• Gaps are also left between sectors to avoid interference between sectors

Disk velocity

• Rotate disk at constant angular velocity (CAV)

• Same number of sectors in different tracks

• Different space between bits in different tracks

• Storage density of the inner ring determines the capacity of disk

• Can use zones to increase capacity

Disk layout methods diagram

Winchester disk

• First disk storage system，305 RAMAC by IBM in 1956, 5MB

• Hard disk 3340 Invented by IBM in 1973

• Has several coaxial metal discs coated with magnetic material

• Disk, magnetic head and drive mechanism are sealed in a box

• The disk has two 30M storage units and the caliber and charge of “Winchester rifle” are also two 30, so it is named

• Magnetic head is not in contact with the disk to improve the durability of the disk

• In 1980, Seagate manufactured the first Winchester hard disk on a personal computer

• Invention of high-sensitivity magnetic head makes it possible for high-density storage

• Almost all mechanical hard disks are based on Winchester technology

• Although the capacity of hard disk has increased many times, the principle is the same as before

Hard disk – Winchester

• Developed by IBM

• Sealed unit

• One or more platters (disks)

• Heads fly on boundary layer of air as disk spins

• Very small head to disk gap

• Getting more robust

Winchester disk format Seagate

Characteristics of disk

• Fixed (rare) or movable head
• Removable or fixed
• Single or double (usually) sided
• Single or multiple platter
• Head mechanism
  • Contact(Floppy)
  • Fixed gap
  • Flying(Winchester)

Floppy disk

• Small capacity

• Slow

• Universal

• Cheap

• Obsolete

Multiple platter

• Install multiple discs in vertical direction

• One head per side

• Heads are joined and aligned

• Aligned tracks on each platter form cylinders

• Data is striped by cylinder

  • Reduces head movement

  • Increases speed (transfer rate)

Cylinders

• Platter has two heads

• Multiple tracks at the same position is called cylinder

• Data is stored according to cylinder rather than sequentially on a certain disk

• Improve the reading and writing speed by reducing the movement of the magnetic head

Typical hard disk drive parameters

Timing of disk I/O transfer

• 读写之前，首先要等待设备和通道空闲

• 寻道时间：磁头从当前位置移动到数据所在的磁道的时间

• 旋转延迟：数据所在的扇区旋转到磁头可以读写的位置

• 数据传输：数据实际的传送阶段

• 存取时间：寻道时间+旋转延迟

• 传输时间：磁道定位后进行数据实际传输的时间

Transfer time ! ! !

• $Transfer\ time - T\newline$

  • $T=\frac{b}{rN}\newline$
  • b = number of bytes to be transferred
  • N = number of bytes on a track
  • r = rotation speed, in revolutions per second

• The total average access time (including transfer time)

• $T_a=T_{total}=T_s+\frac{1}{2}r+\frac{b}{rN}\newline$

• $T_s=seek\ time\newline$

• $\frac{1}{2}r\ is\ the\ average\ Rotational\ latency\newline$

Conclusion

• If the data is stored on the disk completely randomly

  • The read of each sector requires seek time + rotation delay + transmission

  • Most of the time is spent in seek track and sector, which is very time-consuming

• Therefore, data is generally stored in adjacent tracks and sectors in sequence

  • Only one seek time and several sector seeking times are required

  • How data is organized is important

RAID

• Redundant Array of Independent Disks

• Redundant Array of Inexpensive Disks

• Store data in multiple disks to improve I/O performance and increasing the disk‘s capacity

• Parity information was used to prevent equipment failures

Characteristics

• 7 levels in common use

• Not a hierarchy，but 7 schemes

• common characteristics

  • Set of physical disks viewed as single logical drive by OS

  • Data distributed across physical drives

  • Use redundant capacity to store parity information

RAID0

• Can not be considered as a level in RAID

• No redundancy，expanded storage space

• Disks are striped

• Data is stored in the adjacent physical disks in strip order

• Increase speed

  • Multiple data requests probably not on the same disk

  • Disks seek in parallel 磁盘寻道并行化

  • A set of data is likely to be striped across multiple disks

Data mapping for RAID0

Advantages and disadvantages of RAID0

• Advantages

  • I/O performance is greatly improved by spreading the I/O load across many channels and drives

  • Very simple design and easy to implement

• Disadvantages

  • NOT fault-tolerant: the failure of just one drive will result in all data in an array being lost

Applications of RAID0

• Not every application scenario requires high data reliability

• In some scenarios，low cost is more important than reliability

• Supercomputers in which

  • Performance and capacity are primary concerns

• Video production and editing

  • A few data errors do not affect the overall effect

RAID1

• Mirrored Disks

• Data is striped across disks

• 2 copies of each stripe on separate disks

• Read from either

• Write to both

• Recovery is simple

  • Swap faulty disk & re-mirror

  • No down time

• High reliability，expensive

Advantages and disadvantages of RAID1

• Advantages

  • 100% redundancy of data: no rebuild is necessary in case of a disk failure, just a copy to the replacement disk

  • Fault recovery is simple. No down time

  • Simplest implementation

  • May improve read performance

• Disadvantages

  • Expensive!

Applications of RAID1

• Applications： Any applications that request a high reliability

  • Accounting

  • Payroll

  • Financial

  • ……

RAID2

• Using the parallel access technology, all disks participate in the execution of each I/O request. All disks are synchronized

• Very small stripes，often single byte/word

• Calculate error correction code, and then the data and check code are stored on different disks in the disk array

• Hamming code is often used to correct error

• All disks need to be read at the same time

• All disks need to participate in the write, and the error correction code needs to be calculated

• Lots of redundancy

  • Expensive
  • Not often used

RAID3

• Similar to RAID 2，parallel storage and check bit used

• Only one redundant disk, no matter how large the array

• Simple parity bit for each set of corresponding bits

• Data on failed drive can be reconstructed from surviving data and parity info

• Very high transfer rates

RAID4

• Each disk operates independently

• Good for high I/O request rate

• Large stripes

• Bit by bit parity calculated across stripes on each disk

• Parity stored on parity disk

• Update parity when writing data

RAID5

• Like RAID 4

• Parity striped across all disks

• Round robin allocation for parity stripe

• Avoids RAID 4 bottleneck at parity disk

• Commonly used in network servers

RAID6

• For higher fault tolerance

• Two parity calculations

• Stored in separate blocks on different disks

• User requirement of N disks needs N+2

• High data availability

  • Three disks need to fail for data loss

• Significant write penalty

  • Two parity disks need to be updated every time data is written

RAID comparison

Solid State Drives

• A solid state drive is a memory device made with solid state components that can be used as a replacement to a hard disk drive

• A type of EEPROM

• Include interface module, controller, address module, data buffer, error correction module and storage module

Example

• CF、SD、MiniSD、MicroSD

• 固态移动硬盘

• 电脑中的固态硬盘

• U盘

Types of SSD

• Flash Based SSD

  • Flash chip as storage medium, often called SSD

  • Movable

  • No need power supply

  • Commonly used in notebook, micro hard disk, memory card, U disk, etc.

  • Long life and high reliability

    • SLC P/E more than 10000 times

    • MLC can reach more than 3000 times

    • TLC can also reach about 1000 times

    • QLC can also ensure 300 times

• DRAM Based SSD

  • DRAM as storage medium, application range is narrow

  • High performance, theoretically unlimited write

  • Independent power supply is needed

• 3D Xpoint SSD

  • Based on 3D xpoint, close to DRAM

  • Nonvolatile storage

  • Low read delay ,1% of the existing SSD

  • unlimited storage life

  • Density is relatively low and cost is very high

Compare of NOR and NAND

• Flash memory include NOR Flash and NAND FLASH

• NOR

  • Developed by Intel in 1988

  • With dedicated address and data line (similar to SRAM), read and write in byte mode

  • Fast access speed and small storage capacity

  • Data can be read randomly by byte

  • Program can be executed in the NOR

  • Speed of writing and erasing is not fast, which affects its performance

  • Suitable for program storage, such as BIOS

• NAND

  • Reading and writing in blocks

  • Slower in reading, but much faster in writing and erasing than NOR

  • Smaller volume and higher storage density than NOR flash memory

  • Suitable for storing large amount of data

  • Including four types: SLC（Single-Level Cell）,MLC（Multi-Level Cell）,TLC（Trinary-Level Cell）,QLC（Quad-Level Cell）

NAND Flash Storage unit

NAND Flash Diagram

Advantages of SSD

• Read and write fast Continuous R/W over 500MB / s, access time less than 0.1ms (mechanical hard disk generally in 12~14ms)

• Shockproof and fall Resistant

• Low power consumption

• No noise

• Wide working temperature range, -10 ~ 70 ℃ (mechanical hard disk generally in 5 ~ 55 ℃)

• Light

Disadvantages of SSD

• Smaller capacity than mechanical hard disk

• Life limit

  • Limit the number of erasures. SLC about 10000 times, MLC about 3000 times, TLC about 1000 times, QLC only 300 times

  • Through the balanced algorithm to manage the storage unit, reduce the unnecessary amount of writing, improve the service life

• Expensive than mechanical hard disk

• Data storage time is limited. SLC is about 10 years, MLC is shorter

Optical Memory

Optical storage CD-ROM

• CD—Compact Disk

• Invented in 1983

• Originally for audio，later used to store data

• 650Mbytes giving over 70 minutes audio，about 15 songs

• Polycarbonate coated with highly reflective coat, usually aluminium

• Data stored as pits

• Read by reflecting laser

• Surface area of the optical disc is large. In order to store more data, hard disk storage cannot be used

• A spiral that rotates outward from the center to the outermost edge

• Length of the innermost and outermost sectors is the same

• Constant packing density,no loss of capacity

• Constant linear velocity,variable speed rotation

CD-ROM Drive Speeds

• Constant linier velocity, not like hard disk

• Audio is single speed

  • 1.2 m/s

  • Track (spiral) is 5.27km long

  • Gives 4391 seconds = 73.2 minutes

• For data CD, can be rotated in high multiples for faster data reading

  • e.g. 24x

  • Quoted figure is maximum drive can achieve

CD-ROM format

• 数据按照块（扇区）的方式组织在光盘上

• 块的开头是12个字节的同步标识，采用了特殊的字符，开头和结尾的1个字节全为0，中间10个字节全为1

• 同步标识后面是4个字节的块头，包含块（扇区）地址和模式。模式0表示是空的，没有数据。模式1表示后面是2048个字节的数据和288位的纠错码。模式2表示后面是2336个字节的数据，没有纠错码

• 数据域，是用户的数据

Advantages & Disadvantages of CD-ROM

• Large capacity(once)

• Easy to mass produce

• Removable

• Robust

• Expensive for small runs

• Read only

• Slow

Other Optical Storage

• CD-Recordable (CD-R)

  • Disc is coated with a dye layer that can be activated by laser

  • Write only once

  • Now affordable

  • Applicable to document archiving

  • Compatible with CD-ROM drives

• CD-RW

  • Erasable

  • Getting cheaper

  • Can be used as a secondary storage device

  • Mostly CD-ROM drive compatible

DVD

• DVD invented to increase storage capacity

• Digital Video Disk

  • Used to indicate a player for movies

  • Only plays video disks

• Digital Versatile Disk

  • Used to indicate a computer driver

  • Will read computer disks and play video disks

DVD technology

• Very high capacity (4.7G per layer)

• Full length movie on single disk

  • Using MPEG compression

  • Finally standardized (honest!)

  • Movies carry regional coding

  • Players only play correct region films

• Multi-layer，double-side

• Content can be fixed

DVD – writable

• Loads of trouble with standards

• First generation DVD drivers may not read first generation DVD-W disks

• First generation DVD drivers may not read CD-RW disks

• Wait for it to settle down before buying!

High Definition Optical Disks

• Designed for high definition videos

• Much higher capacity than DVD

Magnetic Tape

• Two types of recording

  • Parallel recording

  • Serial recording

• Earlier tapes used nine tracks – parallel recording

• Modern systems use serial recording, referred to as serpentine recording

Characteristics of tape

• Serial access

• Slow

• Very cheap

• Backup and archive

• Linear Tape-Open (LTO) Tape Drives LTO

  • Developed late 1990s

  • Open source alternative to proprietary tape systems