External Memory: Magnetic Disks, RAID, Optical Media & Tape

External Memory: Taxonomy

• Three broad categories emphasized in course (per slide list)
  • Magnetic Disk
  • RAID configurations (0-6)
  • Removable magnetic-disk products (Zip, Jaz, portable USB HDD)
  • Optical
  • CD family (CD-ROM, CD-R, CD-RW)
  • DVD family (press-pressed, ±R, ±RW, multi-layer)
  • Magnetic Tape
  • 8 mm, DAT, DLT and other backup-oriented standards

Hard-Disk Basics & Historical Context

• Invented mid-1950s (IBM 350 in $1954$)
  • Up to $20\;\text{in}$ diameter, only a few MB of capacity
• Naming evolution
  • “Fixed disk” / “Winchester” (IBM codename) → “Hard disk” (to distinguish from flexible “floppy”)
• Magnetic recording principle identical to cassette tapes
  • Medium can be erased, rewritten, and retains flux patterns for many years
• Form-factor & capacity evolution (IBM drives, 15-year chart)
  • Follows exponential trend similar to Moore’s law (graph from $1980$–$2015$ shows jump from $\approx 10^{-3}\;\text{GB}$ to $10^{4}\;\text{GB}$)

Contemporary Desktop Metrics (c. lecture)

• Typical capacity: $10$–$40\;\text{GB}$
• Performance viewpoints
  • Data rate: $\text{bytes}\,/\,\text{s}$ delivered to CPU
  • Seek time: delay between request and first byte available
  • Capacity: total bytes storable

Magnetic Disk Construction

• Substrate material
  • Legacy: aluminium
  • Modern: glass ⇒ improved surface uniformity, lower fly heights, better shock tolerance, fewer defects, reduced read/write errors
• Sealed aluminium housing + detachable controller PCB (“electronics board”)
  • PCB drives spindle-motor and head-servo

Key Components & Terminology

• Platters
  • Rigid disks coated with magnetizable layer
• Spindle + spindle motor
  • All platters clamped to common shaft
  • Typical speeds $3600$–$7200\;\text{RPM}$ (higher in enterprise)
• Read/Write heads
  • One per platter surface; mounted on a shared actuator shaft
  • Heads fly on µm air cushion; rest on landing zone when parked
  • Contamination risk ⇒ manufactured in Class 100/10 clean-rooms
• Head actuator
  • Obsolete stepper-motor designs (prone to heat mis-registration)
  • Present voice-coil servo: coil + permanent magnet, closed-loop positioning
• Advanced head technology
  • Inductive element for writing
  • Magneto-Resistive (MR / GMR / TMR) element for reading ⇒ higher density, separate read/write gap

Beneath the PCB

• Filtered breather hole equalizes pressure
• Arm assembly can sweep hub-to-edge $\approx 50$×/s
• Multiple platters ⇒ multiply storage; e.g., 3 platters → 6 surfaces

Data Organisation & Formatting

• Platters ⇒ concentric circular tracks separated by guard gaps
  • Smaller gap ⇒ higher areal density
• Constant Angular Velocity (CAV) by default
  • Outer track linear velocity higher ⇒ wasted surface; mitigated via ZBR (zoned bit recording) where each zone has fixed bits/track
• Tracks subdivided into sectors (minimum addressable block)
  • PC norm: $512\;\text{B}$ (legacy) or $4096\;\text{B}$ (AF)
• Formatting stages
  • Low-level: writes sector & ECC markers
  • High-level: writes file-system structures (FAT, NTFS, etc.)

Mechanical Timing Parameters

• Seek time (head travel)
• Rotational latency $= \frac{60}{2\times\text{RPM}}$ average
• Access time $= \text{Seek} + \text{Latency}$
• Transfer rate once sector under head

Physical Options

• Head motion: fixed head (1/head/track, rare) vs movable head (1/head/side)
• Portability: fixed vs removable media (e.g., removable HDD cartridges)
• Sides: single- vs double-sided media
• Platter count: single vs multiple (aligned into cylinders)
• Head flight: contact (floppy) vs flying air-bearing (Winchester)

Legacy Floppy Disks

• Form factors & capacities
  • 8″ (IBM, 1971)
  • 5.25″ $360\;\text{kB}$ (IBM PC 1981)
  • 3.5″ $1.44\;\text{MB}$ (mid-1980s onward)
• Named “floppy” due to flexible jacket (5.25″)
• Declined mid-1990s (contamination, low capacity)

Winchester Hard Disk Traits

• Sealed “flying head” design (IBM 3340, 1973, dev. in Winchester, USA)
• Fastest, cheapest /GB bulk external storage; multi-GB commonplace

Removable Magnetic Storage

• Motivations: software distribution, backups, offline security, portability, data sharing
• Zip (Iomega)
  • Higher-grade magnetic coating ⇒ \approx 10× denser tracks
  • Capacities 100\rightarrow250\rightarrow750\;\text{MB}
• Jaz cartridge
  • Enclosed multi-platter HDD sans motor/heads (in drive)
  • GB-class capacity
• External USB HDD
  • Entire sealed 2.5″/3.5″ drive + bridge electronics in portable case

RAID: Motivation & Global Principles

• Enterprises need >\text{TB} scale, high availability, fault tolerance, performance
• RAID = Redundant Array of Inexpensive/Independent Disks (Berkeley, 1987)
• Array appears as one logical volume to OS; data distributed across disks
• Redundancy techniques
  • Mirroring (duplication)
  • Parity (error-coded data)
  • Some levels combine both
• Key benefits
  1. Higher data security (no data loss on single-disk failure)
  2. Fault tolerance (continuous operation)
  3. Improved availability (online rebuild, hot-swap)
  4. Improved performance (parallelism)

Concept 1: Mirroring (RAID 1)

• Data written identically to two drives (duplexing if on separate controllers)
• Pros: 100\% redundancy, near-instant recovery, faster reads (parallel), no rebuild time when one fails
• Cons: 50\% capacity overhead, doubled cost, slower writes (must update both)

Concept 2: Striping

• Splits data across N drives
  • Byte-level (RAID 3, 7) vs block-level (RAID 0, 4, 5, 6)
  • Stripe/Block size selectable at array creation
• Performance: concurrent disk I/O, eliminates single-disk mechanical bottleneck
• No protection unless combined with parity

Concept 3: Parity

• For N$data blocks generate$1$(or$2) parity blocks via XOR/Hamming
  • Any single missing block reconstructable from remaining
  • Distributed vs dedicated parity disk orientation
• Common in RAID 3–6 (RAID 5 most popular enterprise level)

RAID Levels Summary

• RAID 0
  • Block striping, no parity ⇒ max speed, 0 redundancy
• RAID 1
  • Mirroring/duplexing ⇒ full redundancy, high cost
• RAID 2
  • Bit striping + Hamming ECC across multiple parity disks, synchronized spindles ⇒ expensive, unused today
• RAID 3
  • Byte striping + single dedicated parity disk ⇒ high throughput, poor small-I/O (all disks participate)
• RAID 4
  • Block striping + single parity disk ⇒ random-write bottleneck at parity disk
• RAID 5
  • Block striping + distributed parity (round-robin) ⇒ eliminates bottleneck, widespread in servers
• RAID 6
  • Block striping + dual distributed parity ⇒ survives any two drive failures; write-penalty higher; requires N+2 disks

Optical Storage Fundamentals (CD / DVD)

• CDs originally audio (Red Book)
  • Capacity \approx 650\;\text{MB} (74-min audio)
  • Physical stack: polycarbonate substrate 1.2\;\text{mm}, molded pits/lands, reflective Al layer, protective acrylic + label
• Track geometry
  • Single continuous spiral from inner radius outward
  • Pit width 0.5\;\mu\text{m}$, length$\ge 0.83\;\mu\text{m}$, height$125\;\text{nm}
• Player operation
  • Focused laser (780 nm) passes through plastic, reflects; photodiode senses intensity change (pit vs land) ⇒ bits decoded
• Constant Linear Velocity (CLV) rotation
  • Audio single-speed 1.2\;\text{m/s}$linear; spiral length$5.27\;\text{km}$⇒$73.2 min
  • Computer drives quote max “X” multiplier (e.g., 24\times)
• Random access challenges: move sled, adjust RPM, read address, settle
• Pros/Cons
  • Pros: large capacity vs floppy, cheap replication, removable, durable
  • Cons: slow seek/transfer, read-only (for pressed discs), uneconomical for small batches

Recordable & Rewritable CD Variants

• CD-R (WORM): organic dye layer darkens when burnt
• CD-RW: phase-change alloy toggles between amorphous/crystalline (low/high reflectivity)

DVD Technology

• Terminology
  • “Digital Video Disc” (consumer players) versus “Digital Versatile Disc” (computer)
• Capacity boost mechanisms
  1. Smaller pit/track spacing ⇒ higher areal density
  2. Reduced overhead (better ECC)
  3. Multi-layer & double-sided options:
     • Single-sided, single-layer 4.7\;\text{GB}
     • Single-sided, dual-layer 8.5\;\text{GB}
     • Double-sided, dual-layer 17\;\text{GB}
     • Laser wavelength \approx 650\;\text{nm} focus can switch layers (semi-reflective inter-layer)
• MPEG compression allows full-length movies on one disc; region coding (player firmware-enforced, often circumvented)
• Writable DVD chaos (DVD-R/RW, +R/RW, RAM) ⇒ early-generation compatibility issues; market later converged

Magnetic Tape Storage

• Cheapest /GB, ideal for backup & archival; but sequential access ⇒ slow restores
• Drive/media varieties
  • 8 mm (Exabyte): $\approx 6\;\text{MB/s}$
  • DAT (DDS): rotating head (VCR-style), $4\;\text{GB}$ uncompressed, $8\;\text{GB}$ compressed per cartridge
  • DLT (Digital Linear Tape): serpentine head, high reliability & capacity, network-backup workhorse
• Tape summary: serial access, slow, but very low cost, suitable for large unattended backups

Numerical & Formula Highlights

• Rotational latency (average): $\frac{1}{2}\times\frac{60}{\text{RPM}}$
• Access Time: $T<em>{\text{access}} = T</em>{\text{seek}} + T_{\text{latency}}$
• Example RPM values: $3600$, $5400$, $7200$, $10000$, $15000$
• CD spiral length $5.27\;\text{km}$ → $4391\;\text{s}$ audio ($73.2\;\text{min}$)
• DVD layer capacity $4.7\;\text{GB}$ each ⇒ $17\;\text{GB}$ for double-sided dual-layer

Connections & Real-World Relevance

• Clean-room HDD assembly parallels semiconductor fabs (particle contamination kills tolerances $<1\;\mu\text{m}$)
• RAID principles map to cloud-scale distributed file systems (e.g., HDFS replicates blocks akin to mirroring)
• Optical CLV concept reused in Blu-ray (shorter wavelength 405 nm; greater density)
• Tape’s sequential nature mirrors log-structured storage; still dominant in cold-storage warehouses (e.g., LTO-9 $18\;\text{TB}$)

Ethical / Practical Considerations

• Back-up imperative (opening cartoon): data loss vs drive crash
  • Regular backups (RAID ≠ backup) mandatory for resilience
• Disposal of magnetic/optical media should consider data sanitization (degauss, shredding, secure erase) to protect privacy

Example Homework / Further Reading (from slides)

• PC Guide Hard-Disk reference (http://www.pcguide.com/ref/hdd/index.htm)
• HowStuffWorks articles: Hard Disks, Removable Storage, Floppy Drives, SCSI