A developer is creating a web application for a school that stores data in Amazon DynamoDB. The ExamScores table has the following attributes: student_id, subject_name, and top_score. Each item in the

An LSI cannot change the partition key away from student_id; it must share the base table partition key, so a query on subject_name as the partition key is not possible with an LSI.

An LSI with top_score as the partition key is invalid because an LSI must use the same partition key as the base table (student_id), and using top_score as a partition key distributes data poorly.

With subject_name as the GSI partition key, all exam scores for a given subject are grouped together in the index. Using top_score as the sort key means DynamoDB stores them in score order, allowing the application to query the GSI for a subject and retrieve items sorted by top_score efficiently. The student_id attribute is projected into the index and returned in the query results.

A GSI with subject_name as partition key and student_id as sort key groups items by subject but sorts by student_id, not score, making it impossible to efficiently identify the top scorer without reading and sorting all items in application code.