What is the recommended method to avoid unbounded document sizes due to embedding?

The recommended method to avoid unbounded document sizes due to embedding is to break data into multiple collections and use references. MongoDB allows for flexible schema designs, and while embedding data can optimize access times for certain queries, it does come with risks, particularly when documents grow large.

By storing related data in separate collections and using references, you can control the size of any single document, preventing performance degradation and exceeding limits. This approach also helps in maintaining a more modular database design, which can be beneficial for complex applications where data relationships are not strictly hierarchical and can change over time.

In contrast, compressing documents may save storage space but does not resolve the issue of document size limitations during database operations. Data shredding techniques also refer to breaking large documents into smaller parts, but they typically refer to processes that might not be optimal for all use cases. Keeping all data embedded might lead to faster access in the short term, but it increases the risk of hitting document size limits and can complicate updates and maintenance of the data structure.