Message queue concept in HLD - System Design Exercise

Design: Message Queue System

Design the core message queue system including producers, consumers, queue storage, and delivery mechanisms. Out of scope: detailed UI, security/authentication, and complex routing features.

Functional Requirements

FR1: Allow multiple producers to send messages asynchronously

FR2: Allow multiple consumers to receive and process messages

FR3: Ensure messages are delivered at least once

FR4: Support message ordering within a queue

FR5: Handle message persistence to avoid data loss

FR6: Support message acknowledgment from consumers

FR7: Allow scaling to handle 10,000 messages per second

FR8: Provide low latency with p99 latency under 100ms for message delivery

Non-Functional Requirements

NFR1: System must be highly available with 99.9% uptime

NFR2: Messages must not be lost even if a consumer or producer crashes

NFR3: Support horizontal scaling for producers and consumers

NFR4: Latency target: p99 < 100ms for message delivery

NFR5: Throughput target: 10,000 messages per second

Think Before You Design

Questions to Ask

❓ Question 1

❓ Question 2

❓ Question 3

❓ Question 4

❓ Question 5

Key Components

Producers to send messages

Consumers to receive messages

Message queue storage (in-memory or persistent)

Message broker to manage queues and delivery

Acknowledgment and retry mechanism

Load balancer or API gateway for producers and consumers

Design Patterns

Publish-Subscribe pattern

Work Queue pattern

At-least-once delivery with acknowledgments

Message persistence and durability

Horizontal scaling with partitioned queues

Reference Architecture

  +------------+       +--------------+       +------------+
  |  Producer  | ----> | Message Queue| ----> |  Consumer  |
  +------------+       |   Broker     |       +------------+
                       | (Persistent) |
                       +--------------+
                             ^  |
                             |  v
                      +----------------+
                      |  Message Store  |
                      +----------------+

Components

Producer

Any client application or service

Send messages asynchronously to the message queue broker

Message Queue Broker

Custom or open-source broker (e.g., RabbitMQ, Kafka)

Manage queues, store messages persistently, and deliver messages to consumers

Message Store

Durable storage like disk-based database or log storage

Persist messages to prevent data loss on failures

Consumer

Client applications or services

Receive messages from the broker and process them

Acknowledgment Mechanism

Broker feature

Ensure messages are processed and allow retries if needed

Request Flow

1. Producer sends a message to the message queue broker asynchronously.

2. Broker stores the message persistently in the message store.

3. Broker places the message in the appropriate queue.

4. Consumer subscribes to the queue and receives messages.

5. Consumer processes the message and sends acknowledgment to the broker.

6. Broker marks the message as processed and removes it from the queue.

7. If acknowledgment is not received within timeout, broker retries delivery.

Database Schema

Entities: - Message: id (PK), payload, timestamp, status (pending, delivered, acknowledged), queue_id - Queue: id (PK), name, created_at - Consumer: id (PK), name, subscription details Relationships: - Each Message belongs to one Queue (1:N Queue to Message) - Consumers subscribe to Queues (N:N with subscription table)

Scaling Discussion

Bottlenecks

Message broker CPU and memory limits under high throughput

Disk I/O bottleneck for message persistence

Network bandwidth limits between producers, broker, and consumers

Single queue becoming a hotspot causing uneven load

Consumer processing speed limiting message consumption

Solutions

Partition queues to distribute load across multiple broker instances

Use fast persistent storage like SSDs or distributed log storage

Implement load balancing for producers and consumers

Scale consumers horizontally to increase processing capacity

Use batching and compression to reduce network overhead

Interview Tips

Time: Spend 10 minutes clarifying requirements and constraints, 20 minutes designing the architecture and data flow, 10 minutes discussing scaling and trade-offs, and 5 minutes summarizing.

Explain asynchronous communication benefits and use cases

Discuss message durability and acknowledgment for reliability

Describe how to handle ordering and retries

Highlight scaling strategies like partitioning and horizontal scaling

Mention trade-offs between latency, throughput, and consistency