CQRS (Command Query Responsibility Segregation) in HLD - System Design Exercise

Design: CQRS-based System

Design the architecture separating command and query responsibilities including data flow, components, and database schema. Out of scope: detailed UI design and specific technology vendor choices.

Functional Requirements

FR1: Separate the system into command (write) and query (read) parts

FR2: Ensure commands update data consistently

FR3: Allow queries to be optimized for fast reads

FR4: Support eventual consistency between command and query data stores

FR5: Handle 10,000 concurrent users with 70% read and 30% write operations

FR6: API response time for queries should be under 150ms at p99

FR7: System availability should be 99.9%

Non-Functional Requirements

NFR1: Data consistency between command and query sides can be eventual, not immediate

NFR2: Write operations must be strongly consistent

NFR3: Read operations should not block writes

NFR4: System must scale horizontally

NFR5: Latency for command processing should be under 300ms

Think Before You Design

Questions to Ask

❓ Question 1

❓ Question 2

❓ Question 3

❓ Question 4

❓ Question 5

Key Components

API Gateway or Load Balancer

Command Service (handling writes)

Query Service (handling reads)

Command Database (transactional, relational or NoSQL)

Query Database (optimized for fast reads, possibly denormalized)

Event Bus or Message Queue for synchronization

Event Processor to update query database

Cache layer for queries

Design Patterns

Event Sourcing to capture all changes as events

Eventual Consistency for syncing read models

CQRS pattern itself for separation of concerns

Asynchronous messaging for decoupling command and query sides

Reference Architecture

          +-----------------+          
          |   API Gateway   |          
          +--------+--------+          
                   |                   
        +----------+----------+        
        |                     |        
+-------v-------+     +-------v-------+
| Command       |     | Query         |
| Service       |     | Service       |
+-------+-------+     +-------+-------+
        |                     |        
+-------v-------+             |        
| Command DB    |             |        
+-------+-------+             |        
        |                     |        
        |   Event Bus / MQ     |        
        +----------+----------+        
                   |                   
          +--------v--------+          
          | Event Processor |          
          +--------+--------+          
                   |                   
          +--------v--------+          
          | Query DB        |          
          +-----------------+          

Components

API Gateway

Nginx or AWS API Gateway

Routes requests to command or query services based on operation type

Command Service

REST/GraphQL API with transactional support

Handles all write operations ensuring strong consistency

Query Service

REST/GraphQL API optimized for read queries

Handles all read operations optimized for low latency

Command Database

Relational DB like PostgreSQL or NoSQL with ACID support

Stores the authoritative data for writes

Event Bus / Message Queue

Kafka, RabbitMQ, or AWS SNS/SQS

Transports events from command side to query side asynchronously

Event Processor

Microservice or Lambda function

Consumes events and updates the query database accordingly

Query Database

NoSQL DB like Elasticsearch or DynamoDB optimized for reads

Stores denormalized data optimized for fast queries

Cache Layer

Redis or Memcached

Speeds up frequent read queries

Request Flow

1. Client sends a write request to API Gateway

2. API Gateway routes the request to Command Service

3. Command Service validates and writes data to Command Database within a transaction

4. Command Service publishes an event describing the change to the Event Bus

5. Event Processor consumes the event asynchronously from the Event Bus

6. Event Processor updates the Query Database to reflect the change

7. Client sends a read request to API Gateway

8. API Gateway routes the request to Query Service

9. Query Service reads data from Query Database or Cache and returns response

Database Schema

Entities: User, Order, Product Command DB: normalized tables with foreign keys for consistency Query DB: denormalized documents or tables optimized for query patterns Relationships: User 1:N Orders, Order N:M Products via OrderProduct junction Events capture changes like OrderCreated, ProductUpdated

Scaling Discussion

Bottlenecks

Command Database write throughput limits

Event Bus message backlog under high write load

Event Processor lag causing stale query data

Query Database read load spikes

Cache invalidation complexity

Solutions

Scale Command Database vertically or shard by user or entity

Partition Event Bus topics and increase consumer instances

Parallelize Event Processors and use checkpointing

Use read replicas and horizontal scaling for Query Database

Implement cache expiration and event-driven cache invalidation

Interview Tips

Time: Spend 10 minutes understanding requirements and clarifying consistency needs, 15 minutes designing components and data flow, 10 minutes discussing scaling and trade-offs, 10 minutes for Q&A

Explain the separation of command and query responsibilities clearly

Discuss consistency trade-offs and eventual consistency implications

Highlight asynchronous event-driven communication

Mention database choices for transactional writes vs fast reads

Address scaling bottlenecks and solutions realistically