A video surveillance application stores 1 GB average (max 2 GB) files in S3. A Lambda function processes each file once, and processing is I/O intensive requiring multiple reads of the file. After processing the file is deleted. Which option gives the best performance for this workload?

The correct answer is to increase the Lambda function’s ephemeral (/tmp) storage to 2 GB and copy files from S3 into /tmp for processing. Lambda's /tmp directory provides low-latency, high-performance local storage for temporary files, which is ideal for I/O-intensive operations like multiple reads of a large file. Since the files are deleted after processing, this temporary storage is suitable. The maximum file size is 2 GB, so setting /tmp to 2 GB accommodates this. Reading directly from S3 during processing would incur network latency for every read operation, significantly impacting performance for I/O-intensive tasks. Attaching an EBS volume is not possible with Lambda functions. Attaching an Elastic Network Adapter (ENA) is also not a direct option for Lambda functions; ENAs are typically associated with EC2 instances to enhance network performance.