In a document-based database system, such as Couchbase, you don't need to worry about altering tables when there are changes in the schema, as we usually have to do in RDBMS. All schemas of a document are driven by application code only. Whenever there is a need to change schema, we don't need to worry much, as schema of a document goes along with the document itself as metadata. We don't need to change any schema explicitly at the bucket level. We just need to think about the latency, update, and read pattern of documents in terms of dataflow while designing document. We will be more concerned about data flow than data storage. However, my experience leads me to feel, that if you can design a document as a standalone entity, that's the best thing.

We have seen two ways of designing documents: one that includes all related information in one document and others splitting information into multiple documents. If we have to choose the second option, there should be a way...