Design an in-memory database
Overview
This question evaluates understanding of in-memory database engine design, including data model and API surface, in-memory storage layout and indexing, query execution, transaction isolation and concurrency control, durability mechanisms (WAL, snapshots, recovery), and strategies for replication and sharding.
OpenAI
Aug 4, 2025, 10:55 AM
Software Engineer
Technical Screen
System Design
0
0
System Design: In-Memory Database Engine (Single-Node Core, Scale-Out Aware)
Context
Design the core of an in-memory, single-node database engine intended to be the building block for a larger distributed system. Data is primarily memory-resident on each node; durability and scale-out are achieved via logging, snapshots, and replication/sharding across multiple such nodes.
Requirements
- CRUD operations on records.
- Efficient querying: point lookups and range scans; basic filtering and ordering.
- Transaction isolation: define levels supported (at least read committed and snapshot isolation; discuss path to serializable).
- Persistence and backup: crash recovery, snapshots/checkpoints, WAL, and restore.
- Horizontal scalability: strategy for replication and sharding across multiple single-node instances.
Deliverable
Provide a design that covers:
- Data model and API surface.
- In-memory storage layout and indexing.
- Query execution approach.
- Concurrency control and transaction management.
- Durability: WAL, checkpoints/snapshots, recovery.
- Backup/restore strategy.
- Horizontal scaling: replication and sharding.
- Operational concerns: garbage collection, observability, and pitfalls.
Solution
Login required
Show