Skip to main content

Lesson 1: Advanced Patterns

What this lesson covers​

  • Saga pattern for distributed transactions
  • CQRS: Command Query Responsibility Segregation
  • Event sourcing for immutable event stores
  • Idempotency strategies at scale

Read time: 12 minutes

Simple Explanation​

What it is​

These are high-level architectural patterns used when systems are too complex for simple request/response flows.

Why we need it​

At scale, you need patterns that handle distributed data, long workflows, and retries safely.

Benefits​

  • Reliable transactions without locking everything.
  • Scalable reads and writes with CQRS.
  • Auditability through event sourcing.

Tradeoffs​

  • More moving parts and more data to manage.
  • Eventual consistency across services.

Real-world examples (architecture only)​

  • Order saga -> Payment -> Inventory -> Shipping.
  • Event store -> Projections -> Read models.

Saga steps Event sourcing

The Saga Pattern​

Distributed transactions without ACID. Break into steps with compensations:

Saga steps: Saga steps

Pros: Handles long-running transactions, no ACID needed Cons: Complex compensations, eventual consistency

CQRS: Separate Reads and Writes​

Command side (writes): Process events, update primary store Query side (reads): Denormalized views optimized for read patterns

CQRS pattern

Benefit: Independent scaling, optimized views per use case

Event Sourcing​

Store events, not final state. Reconstruct state from events:

Event sourcing

Pros: Complete audit trail, temporal queries Cons: Complex, large event stores, eventual consistency

Idempotency at Scale​

Strategies for idempotent operations:

  1. Idempotency Keys (client-provided unique ID)
# Example request payload
payload = {"amount": 100, "idempotencyKey": "abc123"}
# System deduplicates on key
# Retry safe
  1. Database constraints (unique keys)
INSERT INTO transfers (transferId, amount) VALUES (...)
UNIQUE KEY(transferId)  -- Duplicate inserts fail gracefully
  1. State-based (check if already done)
if order.status == "charged":
  return  # Already processed
charge(order)
order.status = "charged"

What Comes Next​

  • Lesson 2: Distributed systems (CAP, consensus)
  • Lesson 3: Multi-cloud operations

Key Takeaway: Sagas handle distributed transactions. CQRS separates reads and writes. Event sourcing provides audit trails. Idempotency is non-negotiable.