Storage

Computer storage is broadly divided into two categories:

• Primary storage (RAM, cache) - fast but volatile; lost on power off
• Secondary storage (HDDs, SSDs, cloud) - persistent, slower, permanent

---

Common Disk Types

Type	Key Characteristics
SATA	Replaced IDE; 7-pin cable; native hot-swap; seen as SCSI device
SCSI	8-16 bit bus; 5-160 MB/s; multiple versions (Fast, Wide, Ultra)
SAS	Point-to-point serial; better performance than SATA; preferred for servers
USB	Flash drives, floppies; treated as SCSI devices
SSD	No moving parts; lower power; faster transfers; dropping in price; pairs well with rotational drives
IDE/EIDE	Obsolete

Disk Geometry (HDDs)

Rotational disks are made of platters read by heads that scan circular tracks, divided into sectors (traditionally 512 bytes; 4 KB is now standard). A cylinder is the set of the same track across all platters.

Note

Linux uses a logical sector size of 512 bytes for backward compatibility, even on modern 4K-sector drives. SSDs have none of these physical concepts - geometry terms don’t apply to them.

# Inspect physical vs logical sector size
sudo fdisk -l /dev/sdc | grep -i sector
# Sector size (logical/physical): 512 bytes / 4096 bytes

External Storage Architectures

Architecture	Protocol	Characteristics
DAS (Direct Attached Storage)	Block (direct cable)	Fast, reliable, affordable; dedicated to one host; good for small businesses
NAS (Network Attached Storage)	NFS/CIFS over ethernet	Shared file storage; reasonably fast; not suitable for OS installs
SAN (Storage Area Network)	Fibre Channel or iSCSI	Block storage over network; fast, highly available, redundant; expensive

---

Disk Partitioning

Why Partition?

A partition is a physically contiguous section of a disk. Separating storage into partitions provides:

• Isolation: a runaway log file on /var won’t fill up / and crash the OS
• Security: different permissions, quotas, and mount options per partition
• Performance: put frequently-used data on faster media
• Multi-OS support: each OS on its own partition
• Easier backups: separate /home or /data can be snapshotted independently
• Swap isolation: keep swap separate from data

A common minimal layout:

Partition	Purpose
/boot	Kernel and bootloader files (usually 500 MB - 1 GB)
/ (root)	OS and core directories
swap	Virtual memory overflow; also used for hibernation
/home	User files (separate partition for easy re-install)

Tip

Plan partition sizes carefully - resizing after the fact is painful with most filesystem types. LVM can help here (see the LVM page).

Types of Partitions

Type	Description
Primary	Directly bootable; max 4 per disk under MBR
Extended	Container used to hold Logical Partitions; not directly usable as storage; overcomes MBR’s 4-partition limit
Logical	Lives inside an Extended partition; numbered from 5+ (e.g., /dev/sda5)

MBR vs GPT

Feature	MBR	GPT
Max partitions	4 primary (+ logical inside extended)	128 primary (default)
Max disk size	2 TB	8 ZiB (theoretical)
Partition size	Up to 2 TB	Up to 9.4 ZB
Redundancy	Single table in sector 0	Two copies (start + end of disk)
Boot mode	BIOS/Legacy	UEFI (also supports legacy via Protective MBR)

MBR Structure

The MBR occupies the first 512 bytes of a disk:

• Bytes 0-445: boot loader code
• Bytes 446-509: partition table (4 x 16-byte entries)
• Bytes 510-511: magic number 0x55AA (end-of-sector marker)

Each 16-byte partition entry contains: active bit, CHS start/end (ignored by Linux), partition type code, LBA start sector, sector count. Linux uses LBA for all addressing.

GPT Structure

GPT uses a Protective MBR in sector 0 (for backward compatibility), followed by the GPT header and partition entries. The partition table is also copied to the end of the disk, providing redundancy. If the primary header is corrupted, the backup is used to recover.

# Show partition table layout and type
sudo fdisk -l /dev/sda         # MBR or GPT
sudo gdisk -l /dev/sda         # GPT-specific details
sudo parted /dev/sda print     # works with both

Swap Sizing

RAM	Traditional Recommendation
< 2 GB	2x RAM
2-8 GB	Equal to RAM
> 8 GB	At least 4 GB; consider 1x RAM if hibernation needed

Note

Modern systems with large RAM and SSDs often use a swap file instead of a dedicated partition - more flexible and equally fast on SSDs. Some distributions use zram (compressed RAM swap), which is fast but can trigger OOM if overloaded.

---

How Linux Identifies Storage

Device Naming Convention

Linux exposes storage as device nodes under /dev. The naming convention:

/dev/sd[a-z][1-15]       # SATA, SCSI, USB drives
/dev/nvme[0-9]n[1-9]     # NVMe SSDs (namespace-based)
/dev/hd[a-d]             # Old IDE drives (obsolete)
/dev/vd[a-z]             # Virtio virtual drives (VMs)
/dev/mmcblk[0-9]         # SD cards, eMMC

Examples:

Device	Meaning
/dev/sda	First SATA/SCSI disk
/dev/sdb	Second SATA/SCSI disk
/dev/sdb1	First partition on second disk
/dev/sdc4	Fourth partition on third disk
/dev/nvme0n1	First NVMe drive, first namespace
/dev/nvme0n1p1	First partition on above

Block vs Character Devices

Feature	Block Device	Character Device
Access	Random-access, seekable	Sequential, byte-by-byte
Buffering	Can be cached in page cache	Immediate - no kernel buffering
Examples	/dev/sda, /dev/nvme0n1	/dev/tty, /dev/urandom, /dev/null
Major:Minor	Identifies driver type : specific device	Same scheme

Each device has a Major number (identifies the driver) and Minor number (identifies the specific device instance). Visible with ls -l /dev:

ls -l /dev/sda
# brw-rw---- 1 root disk 8, 0 Mar 2 19:00 /dev/sda
#                        ^  ^
#                     major minor

---

Storage Discovery Commands

lsblk - Block Device Tree

lsblk                     # tree view of all block devices and mount points
lsblk -f                  # include filesystem type, UUID, label
lsblk -o NAME,SIZE,TYPE,MOUNTPOINT,FSTYPE,UUID

NAME        MAJ:MIN RM   SIZE RO TYPE  MOUNTPOINTS
sda           8:0    0 465.8G  0 disk
├─sda1        8:1    0     1G  0 part  /boot/efi
├─sda2        8:2    0     1G  0 part  /boot
└─sda3        8:3    0 463.8G  0 part  /home
nvme0n1     259:0    0 476.9G  0 disk
└─nvme0n1p1 259:1    0 476.9G  0 part  /
zram0       251:0    0     8G  0 disk  [SWAP]

blkid - Block Device Attributes

blkid identifies the contents of block devices (filesystem type, UUID, label) by reading metadata fingerprints from the device.

sudo blkid                        # show all devices
sudo blkid /dev/sda1              # specific device
sudo blkid -U "uuid-here"         # find device by UUID
sudo blkid -L "mylabel"           # find device by label
sudo blkid -t TYPE=ext4           # find all ext4 devices

# GPT system example
sudo blkid /dev/sda1
# /dev/sda1: UUID="53ea9807-fd58-4433-9460-d03ec36f73a3" TYPE="ext4"
#   PARTUUID="0c79e35b-e58b-4ce3-bd34-45651b01cf43"

Key attributes:

• UUID: Identifies the filesystem on the partition. Changes when you reformat.
• PARTUUID: Identifies the partition itself (GPT only). Stays same even after reformat.
• LABEL: Human-readable name assigned at format time.

Tip

Always prefer UUID (or PARTUUID/LABEL) over device names in /etc/fstab. Device names like /dev/sda1 can shift after adding/removing drives; UUIDs are stable.

fdisk - Partition Management (MBR/GPT)

sudo fdisk -l                     # list all disks and partitions
sudo fdisk -l /dev/sda            # specific disk
sudo fdisk /dev/sdb               # interactive: create/delete/type partitions

Caution

fdisk writes changes only when you press w. Press q to quit without saving. Always double-check the target device before writing.

Disk Space Overview

df -h                     # disk usage of mounted filesystems (human-readable)
df -Th                    # include filesystem type
df -i                     # show inode usage instead of block usage
du -sh /var/log           # size of a directory
du -sh /* 2>/dev/null | sort -h   # find largest top-level directories