A team is moving to microservices. Each service must only depend on its own datastore. The Payments service (using DynamoDB) needs near-real-time data originating in the Accounts DynamoDB table. What approach gives the simplest, decoupled, and reliable way to obtain updates from Accounts?

DynamoDB Streams provide a near-real-time, ordered stream of item-level modifications in a DynamoDB table. This is the most direct and decoupled way for the Payments service to consume updates from the Accounts table without directly querying it. The Payments service can process these stream records to update its own datastore reliably. Using AWS Glue for frequent ETL jobs introduces latency and complexity for near-real-time requirements. Amazon ElastiCache is a caching service, not a mechanism for propagating database changes between services, and triggers in DynamoDB are not a native feature. Amazon Kinesis Data Firehose is for delivering streaming data to destinations like S3 or Redshift, not directly for propagating DynamoDB changes between tables in a near-real-time, transactional manner.