Discussion on "How to model relational data in DynamoDB"

Stefan Olaru · 2022-03-14T10:18:37.676Z

I'm building monithor.co - an open-source web monitoring service built with Serverless on top of AWS. It's using DynamoDB as the database, a highly performant NoSQL database that's been increasingly popular for serverless applications over the past y...

I wrote a npm package, dyna-record, which abstracts these design concepts. I’m interested to know what you think of it. I’m apparently not allowed to share links here but I wrote a blog title Unlock Relational Data Modeling in DynamoDB with dyna-record on medium

Great to learn something! I'm new to DynamoDB and still confused why you prepended project# and monitor# to some GSIs, this would increase the redundancy and also unintentionally increase storage space.

Also, while querying, why do you query both, the primary key and the GSI, even though the schema says they contain the same data.

I might be totally getting the concept of GSIs wrong. Pardon me for that.

Thanks for the awesome and in-depth article 🎉

Hey Mohit Yadav, good question.

It's called index overloading. The table can hold many different kinds of data items at one time, hence we use generic names for primary key (PK), sort key (SK), and global secondary index (GSI).

Prefixing allows to identify the entity to be queried, e.g. user-related items will always have PK starting with user#, so on.

We can get the project details using GetItem, but to Query all projects of an org we'll have to use the overloaded GSI like GSI=org#1234 AND begins_with(SK, project#).

Thanks for the in-depth answer, Stefan!

Honestly speaking, I understand most of this, but some concepts are still a bit rusty... Explaining them may be out of the scope of this article, so I'll definitely learn them when I get into NoSQL DBs. 😃

Hey :)

Good article explaining the how, but I am left curious about the why :/

Why would you use a dynamoDB for relational data in lieu of just a relational database?

Hey Pablo,

I picked DynamoDB because of the flexible schema and scalability, and it's also very Serverless-friendly. As a plus, the on-demand pricing worked better for my use case.

When a relational database makes more sense, use that by all means.

Stefan Olaru

Ok, thanks for answering :)

Best Article to start with dynamoDB Database design.

Search Hashnode

How to model relational data in DynamoDB

Responses(4)