Question

You are designing a Cosmos DB container for a social media platform. Each user document embeds an array of posts. Product analytics show that active users can accumulate tens of thousands of posts over time, and the document grows continuously. Users frequently read their own profile but only occasionally query all their posts at once. What is the most appropriate data modeling decision?

Accepted Answer

Reference posts as separate items in a posts container partitioned by userId, and keep only a small profile summary embedded in the user document.

Answer

Keep embedding all posts in the user document and rely on autoscale throughput to absorb the larger reads.

Answer

Embed the posts but enable TTL on the user document so old posts are automatically purged.

Answer

Split the posts array across multiple properties in the same document to stay under the size limit.