Design Resilient Architectures

Horizontal scaling adds more instances/nodes to distribute load; vertical scaling increases the size (CPU/RAM) of a single instance. Horizontal scaling improves availability and has near-unlimited headroom; vertical scaling is bounded by max instance size and usually requires downtime/restart.

Horizontal scaling spreads the workload across many instances (ideally across AZs), so the failure of one instance removes only a fraction of capacity. Vertical scaling concentrates the workload on a single larger instance, which remains a single point of failure.

Vertical scaling is capped by the largest available instance type, typically requires a stop/restart (downtime), and does nothing to eliminate the single-instance point of failure. It is simplest for stateful or licensing-constrained workloads that cannot easily distribute.

Loose coupling means components interact through well-defined interfaces (queues, topics, events, load balancers) rather than direct synchronous calls, so producers and consumers can scale, deploy, and fail independently without cascading failures.

A queue buffers messages so a slow or down consumer does not block the producer. The producer keeps writing; the consumer processes at its own rate and retries failed work, absorbing spikes and isolating failures.

Standard SQS provides at-least-once delivery (occasional duplicates) and best-effort ordering (messages may arrive out of order). It offers nearly unlimited throughput. Consumers must be idempotent to handle duplicates.