AWS Integration Pattern Comparison Matrix

Characteristic	REST API	EventBridge	SQS	Kinesis
Maximum Throughput	Depends on API Gateway limits: - 10,000 RPS per account (default) - Can request increase	- 10,000 events/second per account - Can request increase	- Standard: Unlimited - FIFO: 3,000 msgs/sec with batching - 300 msgs/sec without batching	- 1MB/sec or 1,000 records/sec per shard - Unlimited shards
Maximum Payload Size	10MB	256KB	256KB	1MB
Best For	- Request/response patterns - CRUD operations - Synchronous operations - Direct service-to-service communication	- Event routing/fan-out - Service decoupling - Multi-account event routing - SaaS integration	- Decoupled microservices - Job queues - Workload leveling - Ordered processing (FIFO)	- Real-time analytics - Log aggregation - IoT data ingestion - Large scale streaming
Latency	- Milliseconds - Synchronous	- Near real-time - Some latency (seconds)	- Standard: variable - FIFO: ordered - Messages available immediately	- Sub-second - Ordered within shard
Delivery Guarantee	- No built-in retries - Need to implement retry logic	- At-least-once - Retry policy - DLQ support	- At-least-once - Exactly-once (FIFO) - DLQ support	- At-least-once - In-order per shard
Cost Model	- Pay per request - Pay for data transfer	- Pay per event - Pay for custom event bus	- Pay per million requests - Pay for storage	- Pay per shard - Pay for data ingestion/retrieval
Scalability	- Need to request quota increase - Can use regional endpoints	- Automatic scaling - Regional service	- Auto scaling - No shard management	- Manual shard management - Need to split/merge shards
Use When	- Need immediate response - Simple CRUD operations - API-first design	- Complex event routing - Fan-out patterns - AWS service integration	- Need message queuing - Workload decoupling - Need DLQ	- High-throughput streams - Real-time processing - Need ordering
Avoid When	- Asynchronous operations - Need retry mechanisms - Large payload sizes	- Large payloads - Need strict ordering - Ultra-low latency	- Need streaming - Need sub-second processing - Large messages	- Low throughput - Don't need ordering - Cost-sensitive

Additional Considerations:

1. Cross-Region Support

   • REST: Global via CloudFront

   • EventBridge: Cross-region rules

   • SQS: Cross-region queues

   • Kinesis: Cross-region replication

2. Development Complexity

   • REST: Simple to implement

   • EventBridge: Medium (event patterns)

   • SQS: Simple to implement

   • Kinesis: Complex (shard management)

3. Monitoring & Debugging

   • REST: API Gateway metrics/logs

   • EventBridge: CloudWatch metrics/logs

   • SQS: Queue metrics/DLQ

   • Kinesis: Enhanced monitoring