Step 5: Data Storage and Indexing

The data falls into three groups with different access patterns, and each group gets its own store.

Problems. A problem is read by everyone who opens it and edited almost never. A document store such as MongoDB or DynamoDB fits, because one read returns the whole problem as one document. The document holds problem_id, title, description, difficulty, tags, allowed_languages, the time and memory limits, and the visible examples. It also holds test_case_refs, a list of pointers to the hidden test files, but not the files themselves.

problem_id is the primary key