A global e-commerce company wants to centralize customer clickstream data from various regional websites. The data arrives as JSON objects, approximately 1KB each, with a peak ingestion rate of 50,000 records per second. The company needs to transform this data by filtering out bot traffic and enriching it with customer demographic information from a DynamoDB table before storing it in an Amazon S3 data lake in Parquet format. The solution must provide near real-time processing with low latency. Which AWS service is most appropriate for the *transformation* of this data?

Why the other options are wrong

• A. Lambda can process streaming data, but for a continuous stream of 50,000 records/second with complex enrichment and state management, building and managing a Lambda-based solution would be significantly more complex and potentially less cost-effective than a dedicated streaming ETL service like Glue Streaming ETL.
• B. Kinesis Data Firehose offers basic transformations (e.g., format conversion, simple Lambda-based record transformation) but is not designed for complex, stateful, or multi-source enrichment scenarios required here.
• C. Kinesis Data Analytics for Apache Flink is also excellent for real-time processing, but AWS Glue Streaming ETL provides a fully managed, serverless Spark environment which can be more straightforward for ETL-style transformations and writing to S3 in Parquet, especially when integrating with the Glue Data Catalog.