You are viewing a free preview of this lesson.
Subscribe to unlock all 10 lessons in this course and every other course on LearningBro.
When you switch a computer off, the contents of its main memory vanish — yet your files, photos and installed programs are still there when you switch it back on. That permanence comes from secondary storage: non-volatile devices that hold data indefinitely. At A-Level you must understand not merely what the three families of secondary storage are — magnetic (hard disk drives), solid-state/flash (SSDs) and optical (CD, DVD, Blu-ray) — but how each works at a physical level, the performance and durability trade-offs between them, and why a particular technology is chosen for a particular job. This lesson develops all three, with physical-mechanism diagrams, comparison tables, a capacity calculation that links storage to file sizes, and a full specimen question with banded answers.
This lesson addresses the AQA A-Level Computer Science (7517) specification within §4.7 Fundamentals of computer organisation and architecture, specifically the §4.7.1 treatment of secondary storage — the principal types of storage device (magnetic, optical and solid-state), how each operates, and the comparative reasons for choosing one over another.
It links forwards to §4.5.6 (file/data sizes and units of information), since storage capacity is measured in the same powers-of-two units as file sizes, and to the previous input/output lesson, since a drive is a storage peripheral reached through a controller. It builds on the volatility distinction introduced when contrasting RAM with permanent storage.
Main memory (RAM) is volatile — it loses its contents the instant power is removed. Secondary storage is non-volatile: it retains data permanently until deliberately overwritten or erased, so it holds the operating system, applications and user files between sessions.
The trade-off is speed and cost. RAM is far faster than any secondary storage but is expensive per gigabyte and volatile; secondary storage is much cheaper per gigabyte and permanent, but considerably slower. This is why a computer keeps bulk data on disk and copies the parts it is actively using into RAM.
Secondary storage is the bottom layer of a memory hierarchy that every computer uses to reconcile the conflicting demands of speed, capacity and cost. No single technology is fast, huge and cheap all at once, so the machine layers several:
| Level | Technology | Typical access time (order of magnitude) | Capacity | Volatile? |
|---|---|---|---|---|
| Registers | Flip-flops in the CPU | Sub-nanosecond | Bytes | Yes |
| Cache (L1–L3) | Fast SRAM on/near the core | A few nanoseconds | Kilobytes–megabytes | Yes |
| Main memory | DRAM | Tens of nanoseconds | Gigabytes | Yes |
| Secondary storage | SSD / HDD | Microseconds (SSD) to milliseconds (HDD) | Terabytes | No |
Moving down the table, each level gets larger, cheaper per byte and slower. The top three are all volatile working storage; only secondary storage is non-volatile, which is exactly why programs and data must be loaded from secondary storage into RAM to run and saved back to persist. Understanding this hierarchy explains the whole purpose of secondary storage: it is the permanent, capacious, affordable foundation on which the faster volatile layers operate.
An HDD stores data on one or more rapidly spinning platters coated with a magnetic material. Each tiny region of the surface is magnetised in one of two directions to represent a binary 0 or 1. A read/write head on the end of an actuator arm floats microscopically above the surface — on a cushion of air created by the spin, just nanometres from the platter — sensing the magnetic field to read and applying a magnetic field to flip the region's polarity to write. Because the head never touches the surface, a physical shock that makes it strike the platter (a "head crash") can destroy data, which is the root of the HDD's fragility. Capacity has grown over the years chiefly by packing the magnetic regions ever more densely (increasing areal density) and by stacking several platters with a head per surface, rather than by enlarging the disc.
flowchart TB
SP["Spindle (motor axis)"] --> P["Stack of spinning platters<br/>(magnetic discs)"]
AA["Actuator arm"] --> RH["Read / write head<br/>(floats above platter)"]
RH -. "reads/writes via magnetic field" .-> P
| Term | Definition |
|---|---|
| Platter | A circular magnetic disk; data is stored on both surfaces |
| Track | A concentric ring on a platter surface |
| Sector | A segment of a track — the smallest addressable unit (typically 512 bytes or 4 KB) |
| Cylinder | The set of tracks at the same position on all platters |
| Read/write head | Detects or changes the magnetic polarity of the surface |
| Actuator arm | Moves the heads across the platters to the required track |
Because an HDD is mechanical, two physical delays dominate every access:
| Factor | Meaning |
|---|---|
| Seek time | Time for the actuator to move the head to the correct track |
| Rotational latency | Time for the platter to spin so the correct sector is under the head |
| Data transfer rate | Rate at which data streams off the platter once positioned |
| RPM | Revolutions per minute; higher RPM lowers rotational latency. Common values: 5400, 7200, 10000, 15000 RPM |
These mechanical delays are why random access (jumping all over the disk) is far slower on an HDD than sequential access (reading one continuous stretch) — the head must physically reposition for each scattered request.
The link between RPM and rotational latency is a calculable relationship. A drive spinning at 7200 RPM completes:
60 seconds7200 revolutions=120 revolutions per second
so one full revolution takes:
1201 s≈8.3 ms
On average a wanted sector is half a revolution away when the head arrives at the track, so the average rotational latency is about half of that:
28.3≈4.2 ms
The total time to begin reading a randomly placed sector is then roughly the seek time + average rotational latency. If this drive has an average seek time of about 9 ms, a single random access costs on the order of 9+4.2≈13 ms before any data is transferred. By contrast, an SSD reaches the same data in a fraction of a millisecond. This calculation makes vivid why the SSD's advantage is largest for the scattered, random accesses typical of an operating system and applications, and why doubling RPM (which only halves the latency component) cannot close the gap.
| Advantages | Disadvantages |
|---|---|
| Very large capacities (20+ TB) | Moving parts — vulnerable to shock and mechanical failure |
| Lowest cost per GB | Slower than SSDs, especially for random access |
| Mature, well-understood technology | Heavier, bulkier, generates heat and noise |
| Excellent for bulk and archival storage | Higher power draw than an SSD |
An SSD stores data in NAND flash memory: a grid of floating-gate transistors. Each cell traps electric charge in an insulated "floating gate"; the presence or absence of trapped charge represents the stored bit, and because the gate is insulated the charge remains without power — hence non-volatile. Crucially there are no moving parts at all: data is read and written purely electronically, so there is no seek time and no rotational latency.
| Cell state | Floating-gate charge | Stored bit (typical convention) |
|---|---|---|
| Programmed | Charge trapped | 0 |
| Erased | No charge | 1 |
To program a cell, a relatively high voltage forces electrons through the thin insulating layer onto the floating gate, where they stay trapped. To erase it, the voltage is reversed to pull the electrons back off. To read, the controller senses whether the trapped charge is present by testing how the transistor conducts — a low-stress operation, which is why reads are fast and do not wear the cell. It is precisely the repeated forcing of electrons across the insulating layer during programming and erasing that gradually degrades it, giving each cell its finite endurance. Reads, by contrast, are essentially unlimited. This asymmetry — cheap, fast, unlimited reads but costly, wearing writes/erases — shapes everything about how SSDs are managed.
Flash cells are not addressed one at a time. They are grouped into pages (the smallest unit that can be written/programmed, often a few kilobytes) and pages are grouped into blocks (the smallest unit that can be erased, often hundreds of pages). The asymmetry is crucial: you can read and write at page granularity, but you can only erase a whole block at once, and a page must be erased before it can be re-programmed.
This causes the read-modify-write behaviour that explains much about SSDs. To change a few bytes in a page that already holds data, the controller may have to read the surrounding valid pages, erase the whole block, and rewrite everything back. A small logical write can therefore trigger a much larger physical write — an effect called write amplification. It is why SSD writes (especially small, scattered ones) are slower than reads, and why the finite erase budget of each block must be managed carefully.
Manufacturers can store more than one bit per cell by distinguishing several charge levels, trading endurance and speed for capacity and cost:
| Type | Bits per cell | Endurance | Speed | Cost per GB |
|---|---|---|---|---|
| SLC (Single-Level Cell) | 1 | Highest | Fastest | Highest |
| MLC (Multi-Level Cell) | 2 | High | Fast | Moderate |
| TLC (Triple-Level Cell) | 3 | Moderate | Moderate | Lower |
| QLC (Quad-Level Cell) | 4 | Lowest | Slowest of the four | Lowest |
The trend down the table is a single, consistent trade-off: packing more bits into each cell means the controller must distinguish more closely-spaced charge levels, which is slower to read/write reliably and wears out sooner (smaller margins degrade faster), but yields more capacity for the same number of cells, hence lower cost per gigabyte. This is why high-endurance enterprise drives favour SLC/MLC while cheap high-capacity consumer drives use TLC/QLC — the same physics, optimised for different priorities.
A flash cell can be erased and re-programmed only a finite number of times before the insulating layer degrades and the cell wears out. Combined with write amplification, this means naive use would destroy the few blocks holding frequently-changed data long before the rest of the drive. To prevent this, the SSD's controller runs a wear-levelling algorithm: it transparently remaps logical addresses to different physical blocks over time, so that all blocks accumulate roughly the same number of erase cycles rather than concentrating wear on a few. The drive presents the same logical addresses to the OS throughout, but the physical location of any given block of data quietly migrates. Wear levelling, together with spare "over-provisioned" blocks kept in reserve, is what lets a drive with a finite per-cell endurance last for many years of normal use.
| Advantages | Disadvantages |
|---|---|
| Very fast, especially random reads (no mechanical delay) | More expensive per GB than HDDs |
| No moving parts — shock-resistant and silent | Finite write endurance per cell |
| Lower power consumption | Recovery harder if the controller fails |
| Compact and lightweight | Historically smaller capacities (now growing) |
Optical storage occupies a niche the other two cannot: a stamped disc costs pennies to mass-produce yet can be read by a cheap drive anywhere, which made it the dominant medium for distributing music, films and software for decades. Although superseded for everyday storage by SSDs and downloads, the underlying mechanism is still examinable and instructive.
Data sits along a single spiral track as microscopic pits (depressions) and lands (flat areas). A laser shines on the track; pits and lands reflect light differently, and the drive's sensor turns those reflection changes into bits.
| Format | Capacity (single layer) | Laser wavelength | Typical use |
|---|---|---|---|
| CD | ~700 MB | 780 nm (infrared) | Music, small backups |
| DVD | 4.7 GB | 650 nm (red) | Films, medium data |
| Blu-ray | 25 GB | 405 nm (blue-violet) | HD films, large data |
The trend across the three is the key insight: a shorter wavelength laser can be focused to a smaller spot, so it can resolve smaller, more densely packed pits. This is why the capacity rises sharply from CD (780 nm, infrared) through DVD (650 nm, red) to Blu-ray (405 nm, blue-violet) even though all three discs are physically the same size. The shorter the wavelength, the finer the detail the laser can read, so the more bits fit on the disc. Higher capacity is achieved purely by optical resolution, not by making the disc bigger — a point examiners frequently probe. (Capacity is increased further by using multiple recorded layers read at different focal depths, e.g. dual-layer DVDs and Blu-rays.)
Subscribe to continue reading
Get full access to this lesson and all 10 lessons in this course.