5.2a Block device naming: /dev/sda, /dev/nvme0n1 and partition naming conventions¶

When you work with AI infrastructure, you will frequently need to identify and manage storage devices. In Linux, every storage device is represented as a file inside the /dev/ directory. Understanding how these devices are named — like /dev/sda or /dev/nvme0n1 — is essential for configuring storage for AI workloads, such as attaching data drives for training datasets or setting up high-speed NVMe storage for model caching.

⚙️ What Is a Block Device?¶

A block device is any storage device that reads and writes data in fixed-size chunks (blocks). Common examples include: - Hard disk drives (HDDs) - Solid-state drives (SSDs) - NVMe drives - USB flash drives - Virtual disks in cloud or virtualized environments

These devices are accessed through special files in /dev/, and their names follow predictable conventions based on the hardware interface and driver used.

📊 Traditional Naming: /dev/sdX¶

The sd prefix stands for SCSI disk or SATA disk. This naming scheme is used for: - SATA hard drives and SSDs - SCSI drives - USB storage devices - Virtual disks in hypervisors (e.g., VMware, KVM)

How it works:¶

• The first detected SATA/SCSI disk is named /dev/sda
• The second is /dev/sdb, then /dev/sdc, and so on
• Letters continue alphabetically: sda, sdb, sdc, sdd, etc.

Partitions on sdX devices:¶

• The first partition on /dev/sda is named /dev/sda1
• The second partition is /dev/sda2, third is /dev/sda3, etc.
• This applies to all sdX devices

Example naming pattern:¶

• /dev/sda — the entire first SATA/SCSI disk
• /dev/sda1 — first partition on that disk
• /dev/sda2 — second partition
• /dev/sdb — the entire second disk
• /dev/sdb1 — first partition on the second disk

🛠️ Modern Naming: /dev/nvmeXnY¶

NVMe (Non-Volatile Memory Express) drives are much faster than traditional SATA SSDs and are commonly used in AI servers for high-speed data access. Their naming convention is different.

How it works:¶

• nvme indicates the drive uses the NVMe protocol
• The number after nvme is the controller number (starts at 0)
• The number after n is the namespace (usually 1 for a single drive)

Example breakdown:¶

• /dev/nvme0n1 — first NVMe controller, first namespace (the whole drive)
• /dev/nvme1n1 — second NVMe controller, first namespace
• /dev/nvme0n2 — first controller, second namespace (rare, used in multi-tenant setups)

Partitions on NVMe devices:¶

• Partitions are appended with p followed by the partition number
• /dev/nvme0n1p1 — first partition on the first NVMe drive
• /dev/nvme0n1p2 — second partition
• /dev/nvme1n1p1 — first partition on the second NVMe drive

Example naming pattern:¶

• /dev/nvme0n1 — the entire first NVMe drive
• /dev/nvme0n1p1 — first partition on that drive
• /dev/nvme0n1p2 — second partition
• /dev/nvme1n1 — the entire second NVMe drive
• /dev/nvme1n1p1 — first partition on the second NVMe drive

📊 Visual Representation: Linux Block Device Naming Conventions¶

This diagram details the device path naming schemes for SCSI/SATA drives, NVMe drives, and legacy IDE disks in Linux.

🕵️ Comparison Table: sdX vs nvmeX¶

🧩 Partition Naming Conventions Summary¶

🔍 How to Identify Your Devices¶

To see which block devices are present on your system, you can use the following approach:

• Run the lsblk command to list all block devices in a tree format. This will show device names, sizes, and partition layouts.
• The output will display entries like sda, nvme0n1, and their partitions (sda1, nvme0n1p1, etc.).
• Another useful tool is fdisk -l (run with sudo) to list detailed partition tables for all disks.

For reference, a typical lsblk output might look like:

NAME        MAJ:MIN RM   SIZE RO TYPE MOUNTPOINT
sda           8:0    0 465.8G  0 disk
├─sda1        8:1    0   512M  0 part /boot/efi
├─sda2        8:2    0   100G  0 part /
└─sda3        8:3    0 365.3G  0 part /data
nvme0n1     259:0    0   1.8T  0 disk
└─nvme0n1p1 259:1    0   1.8T  0 part /fastdata

📤 Output: The output shows sda with three partitions and nvme0n1 with one partition.

✅ Key Takeaways for New Engineers¶

• /dev/sda is the first SATA/SCSI disk; partitions are /dev/sda1, /dev/sda2, etc.
• /dev/nvme0n1 is the first NVMe drive; partitions are /dev/nvme0n1p1, /dev/nvme0n1p2, etc.
• NVMe drives use a p before the partition number; sdX drives do not.
• Device letters and numbers are assigned in the order the system detects them, which can change after reboots or hardware changes.
• For AI workloads, NVMe drives are preferred for high-speed data access (e.g., loading large models or datasets), while sdX drives are often used for bulk storage.
• Always verify device names using lsblk before performing any operations like formatting or mounting to avoid accidental data loss.