Skip to content

UUID - Universally Unique Identifier

A UUID (Universally Unique Identifier) is a nearly unique 128-bit value that can identify objects across different systems or datasets, functioning like a digital fingerprint, and does not follow a sequential order.

1. UUID Structure

  • 128-bit value represented as 32 hexadecimal characters.
  • Often displayed in five groups: 8-4-4-4-12 with hyphens.
  • Hyphens are optional, included for readability and historical reasons.
  • Example: 67e55044-10b1-426f-9247-bb680e5fe0c8

2. UUID Variants

  • Variant 0: Obsolete
  • Variant 1: Main variant used today
  • Variant 2: Reserved for Microsoft backward compatibility

3. UUID Versions

  • v1: Time and MAC address based; unique but reveals creation time/location.

  • Structure: time_low, time_mid, time_high_and_version, clock_seq_and_version, node

  • v2: POSIX UID replaces low_time; rarely used due to higher collision risk.
  • v3 & v5: Name-based using namespace + name; v3 uses MD5 hashing algorithm, v5 uses SHA1 hashing algorithm.
  • v4: Generated completly randomly and have no identifying information thus small chance of collision.
  • v6: Like v1, but timestamp bits reordered (most significant first).
  • v7: Time-based using Unix Epoch; node replaced with randomness for privacy.
  • v8: Vendor-specific implementations following RFC rules; version indicated in third segment.

A collision occurs when the same UUID is generated more than once and is assigned to different objects.

4. UUID Randomness Capacity

  • A UUID is a 128-bit number.

  • Not all 128 bits are available for randomness:

    • 6 bits are reserved for version + variant.
    • That leaves 122 random bits.

  • So maximum unique random UUIDs = 2×10²² ≈ 5.3×10³⁶

5. Probability of Collision Between Two UUIDs

If we generate two UUIDs, the probability of them colliding is: 1 / 2¹²² ≈ 1.88 × 10⁻³⁷

which is extremly small and therefore can be regarded as practically impossible .

6. How many UUID needs to be generated to get a collision with a certain probability ?

To calculate the number of UUIDs required for a collision with a given probability (p),
The Birthday Bound Formula is used: n = √(2·N·ln(1 / (1 - p)))

For a 10% probability of collision:

n = √(2 · 2¹²² · ln(1 / (1 - 0.1))) ≈ 1,058,482,487,574,424,704 × 10¹⁸ UUIDs

Memory Requirements

Storing all generated UUIDs (128 bits = 16 bytes each):

  • Decimal: 16,935,719.80 TB (≈ 17 million terabytes)
  • Binary (TiB): 15,402,947.43 TiB

Time to Generate

At a rate of 1,000,000 UUIDs per second:

  • Seconds required: ≈ 1.058 × 10¹²
  • Years required: 33,564.26 years

Key Points

  • Collisions are theoretically possible.
  • A 10% chance requires generating about (1.06 × 10¹⁸) UUIDs, needing ~17 million TB of storage and 33,564 years at 1M UUIDs/sec.
  • so Collisions are practically impossible for real systems.

7. UUID Collision Handling

  • Local collisions: If a UUID field is UNIQUE, a duplicate insert will fail and would not be saved. Generating a new UUID resolves the issue.

  • Global collisions: Extremely unlikely; collisions across different systems are virtually impossible.

  • Merging datasets : UUIDs allow multiple sets to combine without remapping. Rare collisions during merging can be detected and handled like local collisions.

8. Integer vs UUID Storage Across Databases

  • Storage size per value:

    • Auto-incrementing integer: 32 bits (4 bytes)
    • UUID (binary): 128 bits (16 bytes)
    • UUID as CHAR(36): 288 bits (36 bytes)

  • Best Practices for Using UUIDs

    • Store UUIDs in a BINARY(16) column instead of CHAR(36) to reduce storage overhead.

9. How to Generate a UUID

Vendor Data Type for UUID Function to Generate UUID
Oracle RAW(16) SYS_GUID()
SQL Server UNIQUEIDENTIFIER NEWID() or NEWSEQUENTIALID()
MySQL BINARY(16) or CHAR(36) UUID() or UUID_TO_BIN(UUID())
PostgreSQL uuid gen_random_uuid() (requires pgcrypto)

10. Use Case of UUID

  • Decentralized generation: UUIDs can be generated anywhere (backend, client, database) and are reliably unique.
  • Simplifies object identity: Maintains uniqueness across disconnected systems.
  • Early ID assignment: Unlike auto-incrementing integers, UUIDs allow determining an object's ID before inserting into a database.

↑ Back to top