Scale a NoSQL Key–Value Store to One Million QPS
Company: LinkedIn
Role: Software Engineer
Category: System Design
Difficulty: medium
Interview Round: Onsite
How would you scale a single-node NoSQL key–value store to support the following workload?
- 50 million key–value pairs.
- 1 million total requests per second.
- A read/write mix of 50% reads and 50% writes.
Explain the distributed architecture, capacity planning, consistency and durability choices, and operational approach. Clarify key/value sizes, traffic skew, latency expectations, and growth before proposing a machine count.
### What a Strong Answer Covers
- Correct interpretation of 500,000 reads and 500,000 writes per second.
- Partitioning, routing, replication, and behavior during failures.
- Capacity estimates based on measured mixed-workload performance and storage/replication overhead.
- Hot keys, rebalancing, and the limits of read caching for a write-heavy workload.
### Follow-up Questions
- What happens if a single key receives a large fraction of the writes?
- How would you move a shard while reads and writes continue?
Overview: Scale a key–value store to 50 million pairs and one million requests per second, balancing reads and writes with sharding, replication, and capacity planning.