Database joins can be expensive and you can reduce dives into the indexes/data by using JSON columns. By loading data kept in smaller 'stub' tables into main tables it is possible to minimize joins for simple data like address, phone numbers, and the like in schemaless JSON documents. Yes, third normal form or better has advantages at many levels but it is a design that is costly for many-to-many relations! It is very easy to refactor data is is often stagnant or infrequently updated to take advantage of hybrid SQL/NoSQL databases to greatly increase overall query performance plus increase code readability. This approach can completely change your ideas on data architecture and provide easy mutability for future changes.

Slides at


Comments are closed.

Scott Dutton at 15:17 on 22 Feb 2019

Really great talk. Very clear examples. Really enjoyed it

Christophe at 17:50 on 22 Feb 2019

I was expecting a bit more comparison and benchmark on how these things can work. Real-life examples would be great.

Also during the slides presenter went from guitars to some student records and then back to guitars which was strange.

Matt Dawkins at 09:49 on 25 Feb 2019

Not quite what I was expecting. For starters, it only filled half the time, which was a bit disappointing. Felt like an announcement that MySQL supports JSON column types, except that that's old news by now, even if we're not all using it. Also disappointing that a talk billing itself on performance gains couldn't give any real-world benchmarks on what those performance gains would be.