Answer: AWS DMS (Database Migration Service)
Most people think DMS is only for migrating databases, but it also supports S3 as a source and Kinesis Data Streams as a target. The key reason it is the correct answer here is that it handles both existing files already sitting in S3 and any new files added going forward — all in one setup. No other option covers both at once without custom code.
Answer: Amazon FSx for Lustre
FSx for Lustre is a high-performance file system built for compute-heavy workloads like machine learning, HPC, video processing, and financial modeling. What makes it special is its native S3 integration — when linked to an S3 bucket, it presents S3 objects as files so your compute can read and write them directly. Active (hot) data lives in Lustre for fast access. Inactive (cold) data is automatically pushed back to S3 at low cost. No other AWS file system offers this combination.
Answer: Golden AMI + EC2 User Data (two things together)
The trick is to split your setup into two parts:
Golden AMI — take everything that never changes (Java runtime, OS packages, libraries, dependencies) and bake it into a custom AMI. When EC2 boots from this AMI, all of that is already installed. No download, no install, instant start.
EC2 User Data — use a boot script for the parts that do change per deployment, like environment variables, config files, or pulling the latest version of your app code.
Together they give you fast boots without sacrificing flexibility.
Answer: Using the primary private IP address of the instance
When you register a target by Instance ID, the NLB does not route using the instance ID itself. It resolves the instance ID to the primary private IP address on the primary network interface of that instance, and then routes traffic there. This is an internal AWS mechanism. The public IP and Elastic IP of the instance are never used in this routing path.