[hox]

Conpletely agreed; using DynamoDB for primary storage suffers from needing to design data models around the underlying technology. This is true for secondary keys as well, and is even constrained by the odd pricing model that DynamoDB follows.

I've been happy using DynamoDB as a large-scale caching layer, but even that only fits very specific use case criteria.

[fragsworth]

> using DynamoDB for primary storage suffers from needing to design data models around the underlying technology.

Every database system, at large scale, suffers from this problem.

[grogenaut]

I laugh every time I see this because everyone seems to forget the rough time they had the first time they encountered a relational database and how to map their problem into that space... Sure it's pretty straight forward now but that's the point... You're still doing it... You are still using your domain knowledge of relational to map your problem ilto the underlying tech. You just don't notice it. "Oh companies have people and people have multiple addresses with one marked as primary and addresses have 1-3 address lines and several optional fields based on local and and and and..." That's several years of learning if you had to discover it yourself there, not just on the address domain but also on how to decompose it.