tune an sqs-to-lambda event source mapping's batching window and concurrency for cost and latency trade-offs

Steps

1. Set BatchSize and MaximumBatchingWindowInSeconds together; batch sizes over 10 require a batching window of at least 1 second.
2. Understand Lambda's default SQS scaling ramp: 5 concurrent batches initially, increasing by up to 300 concurrent invocations per minute up to a maximum of 1,250 concurrent invocations.
3. Set ScalingConfig's MaximumConcurrency to cap concurrent invocations from a queue when downstream systems can't handle full-burst Lambda concurrency.
4. Note that MaximumConcurrency and other throttling approaches are configured per event source mapping and interact with reserved concurrency on the function.
5. Adjust the batching window upward to reduce Lambda invocation counts and lower cost at the expense of added per-message latency, and downward for latency-sensitive workloads.