Intermediate

Pub/Sub Pattern

Messaging & Async

Event-driven architecture with publishers and subscribers decoupled via topics

Core Idea

Publishers emit events to topics; subscribers consume independently. Enables fan-out, selective subscriptions, and loose coupling across services.

When to Use

When multiple services react to the same event, or you want to decouple producers from diverse consumers.

Recognition Cues
Indicators that this pattern might be the right solution
  • Need to fan-out a domain event
  • Multiple teams interested in the same signal
  • Desire to add consumers without changing producers

Pattern Variants & Approaches

Overview
Publishers emit events to a topic; subscribers independently consume with filters and consumer groups.

Overview Architecture

PublishDeliverDeliver⚙️Publisher📬Topic⚙️Subscriber 1⚙️Subscriber 2

When to Use This Variant

  • Fan-out to multiple systems
  • Team autonomy
  • Selective subscriptions

Use Case

Domain events across microservices and near-real-time pipelines.

Advantages

  • Loose coupling
  • Easy fan-out
  • Independent scaling

Implementation Example

# Pub/Sub sketch
topic.publish(event)
sub1.on(event)
sub2.on(event)
Tradeoffs

Pros

  • Loose coupling and team autonomy
  • Easy fan-out and independent scaling
  • Flexible real-time processing

Cons

  • Eventual consistency and debugging complexity
  • Operational overhead (schemas, lag, replay)
  • Hard to guarantee ordering globally
Common Pitfalls
  • Assuming strict ordering across partitions
  • No schema/versioning control causing breakages
  • Backpressure not handled leading to lag
  • Duplicate/out-of-order event handling missing
Design Considerations
  • Push vs pull delivery; durable vs ephemeral subs
  • Topic partitioning and per-subscription filters
  • Schema registry and evolution policies
  • Consumer groups for horizontal scale
  • DLQ and replay policies
Real-World Examples
Apache Kafka

Topic-based pub/sub with consumer groups

Thousands of topics, millions TPS
Google Pub/Sub

Global pub/sub with push/pull

Global multi-region service
AWS SNS

Fan-out to SQS, Lambda, HTTP

Massive event fan-out
Complexity Analysis
Scalability

High - Partitioned topics

Implementation Complexity

Medium - Contracts and lag mgmt

Cost

Low to Medium - Broker/service fees

Related Patterns
Message QueuesEvent SourcingOutboxCQRS