Storage¶

Exercise storage-boot-annex.pdf · PDF

Turn off your phone. Turn it back on. Your photos are still there — but everything that was open is gone. Where did the photos wait while the power was off?

In Computer architecture we called it the pantry: the huge, slow, permanent archive behind the countertop (RAM). This page opens the pantry door. When you buy a laptop, the spec sheet says "512 GB SSD" or "1 TB HDD" — and the difference between those two words decides whether your computer starts in 10 seconds or in a minute.

What is storage?¶

Storage is the part of the computer that keeps data when the power is off.

RAM forgets everything at power off: it is volatile. Storage is non-volatile: your files, your apps and the operating system itself live there, and they survive until you delete them.

In the kitchen: the countertop (RAM) gets wiped every night. The pantry (storage) keeps its shelves full. Before the chef (CPU) can cook with an ingredient, someone has to bring it from the pantry to the countertop — that is what "loading" or "opening" a file means.

Today, three families of pantry fill almost every computer:

Let's open each one.

The hard disk drive (HDD)¶

A hard disk drive (HDD) stores bits as tiny magnetised spots on spinning metal or glass disks.

A little history¶

In 1956 IBM shipped the first commercial hard disk, the IBM 350, part of the RAMAC computer. It had fifty 24-inch platters, weighed about a ton, and stored about 3.75 MB — less than one song in MP3. Today a hard disk that fits in your hand stores more than 20 TB: over five million times more.

How it works¶

Picture a record player, made absurdly precise:

• Platters — rigid disks coated with a magnetic material. Each tiny area is magnetised in one direction or the other: that is a 1 or a 0.
• Spindle motor — spins the platters at a fixed speed, measured in RPM (revolutions per minute): usually 5,400 or 7,200 RPM in desktops and laptops, up to 15,000 in some servers.
• Read/write head — one per platter surface, at the end of an arm that swings across the disk. The head does not touch the platter: it flies on a cushion of air a few nanometres above it. A dust particle is hundreds of times bigger than that gap.

flowchart LR
    CPU["<b>CPU</b><br>the chef asks for<br>a file"] -->|"SATA cable"| CTRL["<b>Controller</b><br>the disk's own<br>small computer"]
    CTRL --> ARM["<b>Arm + head</b><br>swings to the<br>right track"]
    ARM -->|"wait for the sector<br>to spin under the head"| PLAT["<b>Platters</b><br>magnetised spots:<br>the 1s and 0s"]

To read one piece of data, the arm has to move to the right track (seek) and then wait until the right spot spins underneath (rotational latency). Together that takes around 10 ms. That sounds fast — but remember the scale from Computer architecture: if one CPU cycle were one second, an HDD seek would last about a year.

Mechanical also means fragile: dropping a laptop while its HDD is spinning can make the head crash into the platter. That is a head crash, and the data on that spot is gone.

The solid-state drive (SSD)¶

A solid-state drive (SSD) stores bits as electric charges trapped inside chips. Nothing moves — that is what "solid state" means.

Flash memory¶

The chips inside an SSD (and inside your USB stick, your phone and your SD card) are flash memory. It was invented by Fujio Masuoka at Toshiba around 1980; the kind used in SSDs, NAND flash, was presented in 1987. The name came from a colleague: erasing a whole block at once reminded him of a camera flash.

The intuition: flash memory is a grid of billions of microscopic "buckets", called cells. Each cell can trap electrons behind an insulating wall, and they stay trapped with no power. Full or empty bucket = 1 or 0. Modern cells measure several levels of "fullness" to store more than one bit each:

Because there is no arm to move, any piece of data is equally close. An SSD answers in about 0.1 ms — around a hundred times quicker than an HDD's seek.

SATA vs NVMe: same chips, different corridor¶

Here is the trap: two SSDs with the same flash chips can differ in speed by ten times or more. The difference is the corridor — the interface that connects them to the rest of the computer.

• SATA (Serial ATA) was designed in the early 2000s for hard disks. The modern version, SATA III, carries 6 Gbit/s; after overhead, that is about 550–600 MB/s of real data. An SSD plugged into SATA is a race car stuck on a country road.
• NVMe (Non-Volatile Memory Express, version 1.0 in 2011) is a protocol designed from scratch for flash. It skips SATA and talks to the CPU over PCI Express, the same fast lanes the graphics card uses. Most NVMe drives are small sticks plugged into an M.2 slot on the motherboard.

"My 1 TB disk shows 931 GB!"¶

You buy a 1 TB disk, plug it in, and Windows says 931 GB. Did the shop rob you?

No. You met this trap in Binary: there are two rulers.

• The manufacturer uses the decimal ruler: 1 TB = 10¹² bytes = 1,000,000,000,000 bytes.
• Windows uses the binary ruler (2³⁰ bytes per "GB", which is really a GiB) but still writes "GB".

1,000,000,000,000 bytes ÷ 1,073,741,824 bytes per GiB ≈ 931 GiB

Same number of bytes, different unit. Linux and macOS tools often print the unit correctly (931.5G in lsblk means GiB).

Check your own pantry¶

On Linux, one command shows your disks and whether they spin:

lsblk -d -o NAME,SIZE,ROTA,TRAN

NAME     SIZE ROTA TRAN
sda      1.8T    1 sata
nvme0n1  476.9G  0 nvme

• ROTA 1 = rotational: it spins, so it is an HDD.
• ROTA 0 = no moving parts: an SSD.
• TRAN = the corridor: sata, nvme, usb...

The nvme0n1 above was sold as "512 GB": 512 × 10⁹ bytes ≈ 476.9 GiB. The same trap, caught live.

On Windows, open Task Manager → Performance: each disk shows its type (SSD or HDD).

Key takeaways¶

• Storage is non-volatile: it keeps data with the power off. It is the kitchen's pantry; RAM is the countertop.
• An HDD stores bits magnetically on spinning platters; a flying head must seek, which costs ~10 ms per access.
• An SSD stores bits as trapped charges in NAND flash cells: no moving parts, ~100 times faster access, but cells wear out slowly (TBW).
• SATA caps any disk at ~550 MB/s; NVMe uses PCI Express and reaches thousands of MB/s. M.2 is only the shape.
• "1 TB" shows as ~931 GB because the box uses decimal units and Windows uses binary ones.

Quick check¶

Now you know where data sleeps. Next: what happens in the seconds between pressing the power button and seeing the login screen — the boot process.