SQS FIFO vs Standard Queues: Ordering, Dedup, and Throughput

SQS standard and FIFO queues differ on exactly three axes — and the exam hammers all three: message ordering, delivery guarantee, and throughput ceiling.

Standard: throughput over order

• Nearly unlimited TPS — scales automatically with demand
• At-least-once delivery — duplicates WILL happen; consumers must be idempotent
• Best-effort ordering — messages usually arrive in order, not guaranteed
• Cheaper per request and the right default for most decoupling

FIFO: correctness over speed

The subtlety: FIFO ordering is per message GROUP, not queue-wide. Multiple group IDs give you ordered lanes with parallel throughput — high-cardinality group IDs preserve ordering AND scale.

• Exactly-once processing via deduplication (content-based or MessageDeduplicationId)
• Strict ordering within a message group (MessageGroupId) — order across groups is parallel
• Throughput: ~300 msg/s without batching, ~3,000 msg/s with batching — the hard ceiling that decides most questions
• Higher per-request cost than standard

Decision shortcuts

Common trap: choosing FIFO 'to be safe' when ordering isn't required — you pay more and cap throughput for a guarantee nobody needs.

• 'Process payments/events exactly once in order' → FIFO
• 'Decouple microservices, duplicates are handled idempotently' → Standard
• 'Need >3,000 msg/s sustained' → Standard (or shard FIFO by group ID)
• 'Order matters per customer/session, but many independent customers' → FIFO with per-entity group IDs
• 'Just need a buffer between producer and consumer' → Standard