CQRS: Command Query Responsibility Segregation
2 min readSeparate read and write models to scale and optimise each side, and the trade-offs that come with eventual consistency.
Table of Contents
Overview
CQRS stands for Command Query Responsibility Segregation. It separates read (query) and write (command) operations into different models.
“Can be used for situations when app performance and security are complex and objects are exposed to both reading and writing transactions.”
Core Principle
| Model | Responsibility | Optimization |
|---|---|---|
| Write Model (Command) | Handle create, update, delete; manage persistence events; act as data source for read side | Scalability, consistency, event sourcing |
| Read Model (Query) | Generate projections → denormalized views; handle read requests | Performance, fast queries, no joins |
When to Use
- Performance problems — High read volume, model needs frequent updates
- Scalability problems — Large concurrent users, shared model bottleneck
- Flexibility needs — Change read/write independently for new features
Architecture
┌─────────────┐ Commands ┌─────────────┐
│ Client │ ────────────────▶ │ Write Model │
└─────────────┘ └──────┬──────┘
│ Events
▼
┌─────────────┐ Queries ┌─────────────┐
│ Client │ ◀──────────────── │ Read Model │
└─────────────┘ └─────────────┘
▲ │
│ Projections │
└────────────────────────────────┘
Benefits
| Benefit | How Achieved |
|---|---|
| Improved performance | Read model denormalized, optimized for query patterns |
| Increased scalability | Read/write scaled independently |
| Improved flexibility | Change read or write model without affecting the other |
| Security | Different auth/validation per model |
Trade-offs
| Drawback | Mitigation |
|---|---|
| Complexity | Two models to maintain; event synchronization |
| Cost | Double development/maintenance effort |
| Eventual consistency | Read model lags write model (ms to seconds) |
| Not always needed | Only when benefits outweigh costs |
Implementation Patterns
| Pattern | Description |
|---|---|
| Separate tables | Different DB tables for read/write |
| Separate databases | Different DB engines (e.g., write → PostgreSQL, read → Elasticsearch) |
| Event sourcing | Write model emits events; read model rebuilds from event log |
| Materialized views | Read model = precomputed views refreshed on write |
Related
- Saga Pattern — Commands as saga steps
- Transactional Outbox Pattern — Reliable event emission for read model sync