Design a Distributed Unique ID Generator (Snowflake)

Foundations & Core PatternsEasydesign-unique-id-generator

Design a service that hands out 64-bit, roughly time-ordered, globally unique identifiers without a central database sequence becoming a bottleneck.

Snowflake IDsLogical ClocksSequence RangesClock Skew Handling

Traffic & Capacity Estimates:

1M IDs/second across 1,024 nodes · 64-bit IDs · k-sortable by creation time

•Generate 64-bit numeric IDs that are unique across every node in the fleet.
•IDs must be roughly sortable by creation time so they work as primary keys and cursors.
•No coordination round-trip on the hot path — a node generates IDs locally.
•Support node identity assignment when instances are added, restarted, or replaced.

•Generation latency under 1ms at p99, with no network call in the common case.
•Survive clock skew and NTP step-back without ever emitting a duplicate.
•Tolerate a coordination-service outage for the lifetime of an already-leased node ID.

Snowflake layout: 41 bits timestamp (69 years) + 10 bits node ID (1,024 nodes) + 12 bits sequence (4,096 IDs/ms/node).
Theoretical ceiling: 4,096 * 1,000 * 1,024 = ~4.2 billion IDs/second.

Snowflake-style timestamp composition vs UUIDv4 (no coordination, but random ordering destroys B-tree locality) vs UUIDv7.
Node IDs leased from ZooKeeper/etcd vs derived from host metadata — leasing costs a dependency, derivation risks collisions.
Clock moved backwards: refuse to generate and block, or borrow from the sequence bits — availability vs monotonicity.

Click or drag a component onto the canvas, then connect the handles to draw the data flow.

3 nodes · 2 edges

Components · 35

Client & Edge4

Compute & Gateway7

Storage & Caching11

Messaging & Streaming6

Coordination & Ops5

Intelligence2