Design a Column-Queryable KV Database
Company: Snapchat
Role: Software Engineer
Category: System Design
Interview Round: Onsite
Design a distributed key-value database for sparse records, where each row has a primary key and a flexible set of named columns. The system must support:
- `put(rowKey, columnName, value, timestamp)`
- `get(rowKey, columnName)`
- `getRow(rowKey)`
- `queryByColumn(columnName, predicate)` for equality and range filters, such as all rows with `status = active` or `age between 20 and 30`
Assume billions of rows, high write throughput, low-latency point reads, and moderate secondary-index queries. Explain the data model, storage engine, indexing strategy for column queries, sharding, replication, consistency model, and how you would handle schema evolution and hot partitions.
Quick Answer: This question evaluates understanding of distributed storage architectures, secondary indexing and queryability for sparse, column-oriented key-value stores, including competencies in data modeling, sharding, replication, consistency, and schema evolution.