Product design question

Create a scalable system that creates unique IDs.

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Distributed systems design fundamentals for a classic, high-throughput unique-ID generation problem.

Clarify requirements: throughput, uniqueness guarantee across distributed nodes, and whether IDs need to be sortable or time-ordered.
Reject a single centralized counter as the sole mechanism, since it becomes a bottleneck and single point of failure at scale.
Propose a Snowflake-style approach: composite IDs of timestamp, machine or node ID, and a local sequence number, generated independently with no central coordination.
Address clock skew: use monotonic or NTP-synced clocks with a wait-for-clock mechanism to avoid collisions if the clock moves backward.
Address node ID assignment: a lightweight coordination service to assign unique machine IDs and avoid collisions across data centers.
Define success as near-zero collision rate at target throughput and low generation latency per ID.

Names a concrete, industry-standard pattern, Snowflake IDs, rather than inventing an ad hoc scheme, showing real systems knowledge.
Explicitly addresses the clock-skew edge case, a common failure point interviewers probe for.
Distinguishes the need for sortability, which changes the design between random and time-ordered IDs.

Proposing a single centralized ID generator with no discussion of it being a bottleneck or single point of failure.
Ignoring clock skew as a real failure mode in a timestamp-based scheme.

Chapters of the AI PM course that teach what this question tests.