Mock system design interview at Snowflake
A live 45-minute round on a question reported in Snowflake interviews, run by an AI interviewer that asks one thing at a time, probes what you say, looks at your whiteboard, and weighs what that round is reported to weigh. It ends with a debrief: a score, what was strong, what was missing.
Design a Distributed Key-Value Store
A Dynamo-style store that keeps 10 billion keys on dozens of machines, survives node and rack failures, and lets each caller trade consistency for latency with N, R and W.
You will not see the requirements, the numbers or the design before you start. That is the point.
What the Snowflake round is reported to look like
About an hour, and more often infrastructure-shaped than product-shaped. Reported prompts include a key-value store with versions, a quota shared by several products that must never be overdrawn, a job scheduler and a web crawler.
What it weighs
- Storage internals: how data is laid out, indexed and partitioned for the workload
- Correctness under contention, such as concurrent requests drawing on the same quota
- The workload behind each choice, such as fast inserts against fast lookups, with the tradeoff stated
The common mistake
Bringing a consumer-app template. Prompts are reported to be data and storage problems far more often than a social feed, so expect to be asked how records are stored, indexed and kept consistent, not only which services talk to which.
From public interview guides and candidate reports, and Snowflake’s own interview guidance where it publishes any. Not endorsed by Snowflake. The interviewer is an AI running in that style; it is practice, not a prediction.
Start the round
The interviewer’s opening question is free. Answering it is part of Pro. See an example round and its debrief →
The 3 questions this round draws from
Each is reported in Snowflake interviews. Pick one to sit, or read its reference design and walk through a strong answer first.
- Design a Distributed Key-Value StoreA Dynamo-style store that keeps 10 billion keys on dozens of machines, survives node and rack failures, and lets each caller trade consistency for latency with N, R and W.drawn
- Design Apache AirflowSchedule and run twenty thousand data pipelines as dependency graphs, each run tied to a slice of time, so a failed night can be retried or a year backfilled safely.
- Design a Vector DatabaseA managed vector database: 10,000 tenants, 20 B vectors, filtered top-k in under 50 ms at 0.95 recall, and writes searchable in seconds.
Other companies: Anthropic · Google · Meta · Amazon · OpenAI · Microsoft · Uber · DoorDash