Configure SQS FIFO high-throughput mode and choose deduplication scope for predictable ordering at scale

Steps

1. Enable high-throughput mode on a FIFO queue at creation or via SetQueueAttributes: set FifoThroughputLimit=perMessageGroupId and DeduplicationScope=messageGroup
2. With perMessageGroupId throughput limit, each message group ID can sustain up to 3,000 messages per second with batching (or 300 TPS without batching); total queue throughput scales with the number of distinct message group IDs in use
3. Set ContentBasedDeduplication=true to automatically derive the deduplication ID from the SHA-256 hash of the message body, or provide an explicit MessageDeduplicationId on each send for deterministic control
4. Design message group IDs for parallelism: use a fine-grained key (e.g. order_id) to maximise throughput; coarse keys (e.g. a single static group) collapse all traffic to one sequenced lane
5. Test ordering guarantee: within a single message group ID, messages are delivered strictly in-order and only one consumer processes messages from a group at a time; across groups, ordering is not guaranteed
6. For deduplication scope=queue (the classic non-high-throughput setting), a duplicate check applies queue-wide within the 5-minute deduplication window — this is incompatible with high-throughput mode