A company has an application that generates a weather forecast that is updated every 15 minutes with an output resolution of 1 billion unique positions, each approximately 20 bytes in size (20 Gigabytes per forecast). Every hour, the forecast data is globally accessed approximately 5 million times (1,400 requests per second), and up to 10 times more during weather events. The forecast data is overwritten every update. Users of the current weather forecast application expect responses to queries to be returned in less than two seconds for each request.
Which design meets the required request rate and response time?

A. Store forecast locations in an Amazon ES cluster. Use an Amazon CloudFront distribution targeting an Amazon API Gateway endpoint with AWS Lambda functions responding to queries as the origin. Enable API caching on the API Gateway stage with a cache-control timeout set for 15 minutes.
B. Store forecast locations in an Amazon EFS volume. Create an Amazon CloudFront distribution that targets an Elastic Load Balancing group of an Auto Scaling fleet of Amazon EC2 instances that have mounted the Amazon EFS volume. Set the cache-control timeout for 15 minutes in the CloudFront distribution.
C. Store forecast locations in an Amazon ES cluster. Use an Amazon CloudFront distribution targeting an API Gateway endpoint with AWS Lambda functions responding to queries as the origin. Create an Amazon Lambda@Edge function that caches the data locally at edge locations for 15 minutes.
D. Store forecast locations in Amazon S3 as individual objects. Create an Amazon CloudFront distribution targeting an Elastic Load Balancing group of an Auto Scaling fleet of EC2 instances, querying the origin of the S3 object. Set the cache-control timeout for 15 minutes in the CloudFront distribution.