Physical Structure

Before we can understand how databases store, retrieve, and organize data, we must first understand the physical medium upon which all persistent data ultimately resides. Though solid-state drives (SSDs) are increasingly common, magnetic hard disk drives (HDDs) remain the dominant storage technology for large-scale database systems due to their cost-effectiveness and capacity advantages.

Why does physical structure matter for databases?

Unlike main memory (RAM), where any byte can be accessed in approximately the same time, disk access is governed by the physics of mechanical motion. The disk head must physically move to the correct location, and the platters must rotate to bring the desired data under the head. These mechanical constraints create access time asymmetries that are orders of magnitude larger than electronic memory access.

This fundamental reality shapes virtually every aspect of database design:

• Why databases read and write in fixed-size pages rather than individual bytes
• Why sequential access is vastly preferred over random access
• Why buffer pools and caching are essential for performance
• Why index structures like B+-trees are designed to minimize disk seeks
• Why query optimizers estimate costs in terms of disk I/O operations

A hard disk drive is a marvel of precision engineering, consisting of multiple mechanical and electronic components working in concert to store and retrieve data. Understanding each component's role is essential for comprehending disk performance characteristics.

The Primary Components:

At its core, an HDD consists of:

1. Platters — Circular disks coated with magnetic material where data is actually stored
2. Spindle — The central axis around which platters rotate
3. Read/Write Heads — Electromagnetic devices that read and write magnetic patterns
4. Actuator Arm — The mechanical arm that positions the heads over the correct location
5. Actuator Motor — Moves the arm assembly across the platter surface
6. Controller Board — Electronics that manage disk operations and interface with the host system

Hard disk drives operate in a hermetically sealed enclosure to protect the extremely sensitive internal components from contamination. This sealed environment is critical for reliable operation.

Why Sealing is Essential:

The gap between the read/write head and the platter surface is so small that even microscopic particles would cause catastrophic damage. A single smoke particle (approximately 250 nanometers) is nearly 100 times larger than the head-to-platter gap. A fingerprint on a platter would be like placing a mountain range in the path of the flying head.

The Breathing Mechanism:

Despite being sealed, traditional HDDs are not fully airtight. They include a breather hole with a filter that allows air pressure to equalize while blocking contaminants. This is necessary because:

• Internal air pressure changes with temperature
• The actuator arm movements create pressure variations
• Air bearings require controlled airflow

Helium-Filled Drives:

Modern high-capacity drives (8TB+) often use helium instead of air inside the enclosure. Helium offers several advantages:

• Lower density — Reduced turbulence allows heads to fly closer to platters
• Less drag — Lower power consumption and heat generation
• More platters — Thinner gaps allow more platters in the same form factor
• Better stability — Reduced vibration improves read/write accuracy

The platters are the actual storage medium—circular disks where magnetic patterns encode binary data. Understanding platter construction explains capacity limitations and reliability characteristics.

Platter Substrate:

The base material of a platter (the substrate) has evolved over time:

• Early drives (1950s-1990s): Aluminum alloy platters
• Modern drives: Glass or ceramic substrates

Glass/Ceramic Advantages:

• Superior surface smoothness (enabling smaller flying heights)
• Better thermal stability (less expansion/contraction with temperature)
• Higher rigidity (reduced flutter at high rotation speeds)
• Resistance to damage during manufacturing

The Magnetic Layer Stack:

On top of the substrate, several layers are deposited:

1. Underlayer — Promotes adhesion and proper magnetic properties
2. Magnetic Recording Layer — Where data is actually stored (typically cobalt-based alloys)
3. Protective Overcoat — Carbon-based layer protecting against head contact
4. Lubricant Layer — Ultra-thin lubrication (typically 1 nanometer thick)

Magnetic Recording Methods:

The orientation of magnetic domains on the platter determines the stored data. Modern drives use one of several recording technologies:

• Longitudinal Recording (legacy) — Magnetic domains aligned parallel to platter surface; limited density
• Perpendicular Magnetic Recording (PMR) — Domains aligned perpendicular to surface; higher density
• Shingled Magnetic Recording (SMR) — Overlapping tracks like roof shingles; highest density but complex write behavior
• Heat-Assisted Magnetic Recording (HAMR) — Uses laser heating to enable smaller magnetic grains; emerging technology

Hard disk drives come in standardized physical form factors that define their external dimensions and mounting configurations. Understanding these form factors is essential for system design and capacity planning.

Standard Form Factors:

• 3.5-inch — Desktop and enterprise storage; highest capacity; lower cost per GB
• 2.5-inch — Laptops and dense server arrays; lower power consumption
• 1.8-inch — Ultra-portable devices (largely obsolete, replaced by SSDs)

Key Dimensional Specifications:

The form factor name refers to the approximate diameter of the platters. The actual physical dimensions are standardized by industry specifications.

Multi-Platter Configurations:

Modern high-capacity drives contain multiple platters stacked on a single spindle, each with two recording surfaces (top and bottom). The number of platters directly impacts:

• Total Capacity — More surfaces = more storage space
• Heat Generation — More platters = more heat from friction and motor work
• Reliability — More heads increase failure probability
• Access Complexity — Head switching occurs between surfaces

Example Configuration:

A 20TB enterprise drive might contain:

• 9 platters with 18 recording surfaces
• Each surface storing approximately 1.1TB
• 18 read/write heads (one per surface)
• Running at 7200 RPM

The spindle assembly is the rotating core of the hard disk drive, responsible for spinning the platters at precise, constant speeds. The spindle motor's characteristics directly determine rotational latency—one of the key components of disk access time.

Spindle Motor Types:

Modern drives use brushless DC motors with either:

• Ball Bearing Motors — Traditional design; affordable but generates more noise and vibration
• Fluid Dynamic Bearing (FDB) Motors — Modern design using oil film instead of metal contact; quieter, more reliable, less vibration

Rotation Speeds (RPM):

The rotation speed is measured in Revolutions Per Minute (RPM) and is a critical performance specification:

• 5400 RPM — Energy efficient; quiet; lower performance; consumer-grade
• 7200 RPM — Standard desktop and enterprise drives; balanced performance
• 10000 RPM — High-performance enterprise; reduced latency
• 15000 RPM — Maximum performance; highest cost; used in mission-critical systems

The Physics of Rotation:

The spindle motor must maintain precise rotational speed despite varying loads and environmental conditions. Key engineering challenges include:

• Speed Stability — Variations in RPM cause read/write errors; controllers use phase-locked loops to maintain exact speed
• Startup Time — Platters must reach operating speed before data access; enterprise drives typically spin up in 10-30 seconds
• Heat Management — Motor generates heat proportional to RPM²; higher speeds require better cooling
• Vibration Control — Unbalanced rotation causes resonance; precision manufacturing and active damping are essential

Linear Velocity Variation:

An important consequence of rotational motion is that the linear velocity of the platter surface varies with radius. The outer edge moves faster than the inner edge:

• At 7200 RPM, outer track linear velocity: ~120 km/h
• At 7200 RPM, inner track linear velocity: ~60 km/h

This velocity difference affects data density and is addressed by zone bit recording (discussed in subsequent pages).

The read/write assembly consists of the actuator arm (also called the head stack assembly), the read/write heads, and the voice coil motor that positions them. This assembly is responsible for seek time—typically the largest component of disk access latency.

The Actuator Arm:

The actuator arm is a precision-engineered boom that positions the read/write heads over the correct track. Key characteristics include:

• Pivoting Design — Rotates around a pivot point (like a tone arm on a record player)
• Rigid Construction — Must maintain precise positioning under acceleration
• Multiple Arms — One arm per platter surface, all mechanically connected
• Head Stack Assembly (HSA) — The complete assembly of arms and heads moving as one unit

Voice Coil Motor (VCM):

The actuator is positioned by a voice coil motor—the same technology used in loudspeakers. A coil of wire moves through a magnetic field, and current through the coil controls position:

• Proportional Control — Current magnitude controls acceleration
• Bidirectional — Polarity controls direction
• No Gears — Direct drive eliminates mechanical wear
• Microsecond Response — Enables rapid repositioning

The Read/Write Heads:

Modern read/write heads are thin-film inductive write / magnetoresistive read heads mounted on sliders that 'fly' above the platter surface on an air bearing:

• Air Bearing Surface (ABS) — Precisely machined surface that generates aerodynamic lift
• Flying Height — Gap between head and platter (approximately 3-5 nanometers)
• Giant Magnetoresistive (GMR) Read Element — Detects magnetic field changes as resistance changes
• Inductive Write Element — Creates magnetic fields to flip domain orientation

Head Positioning Servo System:

To maintain track positioning with nanometer-level accuracy, drives use a closed-loop servo system:

1. Servo Data — Pre-written positioning information on every track
2. Read Head — Reads servo patterns while seeking and tracking
3. Controller — Calculates position error from servo signal
4. Voice Coil Current — Adjusted in real-time to minimize error

This servo loop operates at sampling rates of 40,000+ times per second, constantly adjusting head position to stay centered on the track.

The controller board (printed circuit board attached to the HDD enclosure) contains all the electronics that transform raw magnetic signals into usable data and manage disk operations. This 'brain' of the drive handles sophisticated signal processing, error correction, and caching.

Key Controller Components:

The Drive Cache (Buffer Memory):

Modern drives include 32-256+ MB of DRAM cache that dramatically improves performance:

• Read Caching — Recently read sectors are cached; re-reads served from memory
• Read-Ahead — Controller speculatively reads sectors following the requested location
• Write Caching — Writes are acknowledged before reaching magnetic media (risking data loss on power failure)
• Command Queueing — Multiple pending commands allow the drive to optimize seek order

Error Correction Coding (ECC):

Raw magnetic media has non-negligible error rates. The ECC engine adds redundancy and can recover from:

• Random Bit Errors — Transient reading errors from noise
• Burst Errors — Multiple consecutive bit errors from media defects
• Sector-Level Errors — Severe errors that require stronger correction

Modern drives use sophisticated codes like LDPC (Low-Density Parity-Check) that add approximately 5-10% overhead but provide extremely strong error correction.