Considerations for Alt-Relational Modeling

Alt-relational modeling chapter conclusion

This is the alt-relational modeling chapter conclusion. I might add more considerations here as they come up.

There are a couple of examples I’m still either working on or having reviewed. Expect those at some point in the near future.

Up next are articles on analytical data modeling, the levels of data modeling (conceptual, logical, and physical), and more. I’ve got around 200 pages of unpublished thoughts on working with people and process, so expect some of that trickling out as well.

As always, comment where you wish. If you find errors, please submit them here.

Thanks,

Joe Reis¹

---

When choosing alt-relational modeling, some confusion and questions are warranted. Are you better off sticking with relational modeling? If not, what approach should you choose and why? What’s with this “schemaless” stuff, anyway? Finally, how does alt-relational work if you migrate part or all of your data model to and from alt-relational? Let’s briefly explore each of these points.

“What Approach Should I Choose?”

A question I often hear about alt-relational modeling is, ”If it’s that easy to use, why would I use the relational model?” This is the classic case of choosing the right tool for the right job. We studied the relational model because it provides a great starting point for thinking about how your data is modeled. When your data is free of redundancies and update/delete side-effects, it's immediately of better quality and utility. You can always choose to denormalize to add redundancy to your data model.

Some say the relational model is brittle because it is strict and rigid. I disagree and think the opposite. The relational model is robust and strict. A loosely defined data model is brittle because seemingly anything can contaminate it. Do you want to introduce redundant data that might cause query problems? No problem. However, while the relational model is robust regarding data integrity, it’s notoriously inflexible for rapidly changing schema or constant iterations to the data model.