Structuring document data.
WHAT IT IS
A document database, with collections containing documents, which may contain subcollections.
WHAT DIFFERS FROM RELATIONAL THINKING
No joins Queries must be supported by an index Structure follows how you read, not how you store
WHAT THAT MEANS
Model for your queries, not for normalisation.
WHAT DENORMALISATION MEANS HERE
Duplicating data so a screen can be rendered from one read.
WHY THAT IS ACCEPTABLE
Reads are billed, and joins do not exist, so fewer reads is the objective.
WHAT IT COSTS
Keeping duplicates consistent when the source changes.
WHAT TO USE FOR THAT
Server-side functions updating duplicates on change.
WHAT TO AVOID
Very large documents Collections where you must read everything to answer a question Unbounded arrays within documents
WHAT TO KNOW ABOUT BILLING
You are billed per document read, written and deleted.
WHAT THAT MEANS
A screen reading hundreds of documents costs accordingly, every time it opens.
WHAT TO DESIGN FOR
Few reads per screen.
WHAT TO USE FOR COUNTS AND AGGREGATES
Maintained counters, or the aggregation queries where they suffice.
