Altitude 4: Messaging & Scale

Altitude 4 — Messaging & Scale

Primary sources: Azure Cloud Design Patterns (https://learn.microsoft.com/azure/architecture/patterns/) and Hohpe, Gregor; Woolf, Bobby. Enterprise Integration Patterns (EIP). Addison-Wesley, 2003. https://www.enterpriseintegrationpatterns.com/

• Queue-Based Load Levelling — Azure Cloud Design Patterns. https://learn.microsoft.com/azure/architecture/patterns/queue-based-load-leveling Absorb the lunch/dinner rush.
• Competing Consumers — Azure Cloud Design Patterns (https://learn.microsoft.com/azure/architecture/patterns/competing-consumers); EIP. A pool of workers draining the courier-assignment queue.
• Publish/Subscribe — Azure Cloud Design Patterns (https://learn.microsoft.com/azure/architecture/patterns/publisher-subscriber); EIP Publish-Subscribe Channel. OrderPlaced fans out to kitchen, courier, customer, and analytics.
• Transactional Outbox — Chris Richardson, microservices.io. https://microservices.io/patterns/data/transactional-outbox.html The reliable bridge between the order's database transaction and the broker.
• Dead-Letter Queue / Backpressure — EIP Dead Letter Channel; Reactive Streams backpressure (https://www.reactive-streams.org/). Where an unprocessable order event lands.
• Saga / Process Manager — Hector Garcia-Molina & Kenneth Salem, "Sagas," ACM SIGMOD, 1987 (https://www.cs.cornell.edu/andru/cs711/2002fa/reading/sagas.pdf); Chris Richardson, microservices.io (https://microservices.io/patterns/data/saga.html). Order fulfilment as charge → assign courier → confirm with restaurant, with compensations (refund) on failure. Heavy skip-if.
• Claim-Check — EIP (https://www.enterpriseintegrationpatterns.com/patterns/messaging/StoreInLibrary.html); Azure Cloud Design Patterns (https://learn.microsoft.com/azure/architecture/patterns/claim-check). Store the large receipt, pass a token.
• Honorable mentions — Idempotent Consumer (see Altitude 5); Event-Carried State Transfer (Fowler, https://martinfowler.com/articles/201701-event-driven.html); Priority Queue (Azure Cloud Design Patterns).
• Brokers (neutral) — GCP Pub/Sub · AWS SQS + SNS · Azure Service Bus.

Altitude 5 — Resilience

Primary source: Nygard, Michael T. Release It! Design and Deploy Production-Ready Software. 2nd ed., Pragmatic Bookshelf, 2018. Library: Polly (https://www.pollydocs.org/).

• Circuit Breaker — Nygard, Release It!; Fowler bliki. https://martinfowler.com/bliki/CircuitBreaker.html Stop calling a failing payment provider.
• Bulkhead — Nygard, Release It! Isolate surge-pricing calls from order placement so one can't sink the other.
• Timeout — Nygard, Release It! Bound the wait on the payment gateway.
• Steady State — Nygard, Release It! Bound every growing resource: purge old courier GPS pings, cap the courier-location cache, rotate logs. Honorable mentions here: Graceful Degradation, Load Shedding, Failover / Redundancy.
• Retry with exponential backoff + jitter — Marc Brooker, AWS Architecture Blog, "Exponential Backoff And Jitter." https://aws.amazon.com/blogs/architecture/exponential-backoff-and-jitter/ Retry a flaky payment authorisation without stampeding.
• Rate Limiting / Throttling — Azure Cloud Design Patterns (https://learn.microsoft.com/azure/architecture/patterns/throttling); .NET rate limiting (System.Threading.RateLimiting). Cap the public order API.
• Idempotency — idempotency keys, Stripe API documentation (https://stripe.com/docs/api/idempotent_requests; illustrative). The property that makes a retried order placement safe: never double-charge, never double-place.

The compounding-failure illustration (95% per-step success over N steps = 0.95^N, roughly 60% at ten steps and 36% at twenty) is arithmetic, not a measured benchmark. Treat it as a worked example, never as a cited statistic.